1
0
mirror of http://galexander.org/git/simplesshd.git synced 2024-11-16 03:59:03 +00:00
simplesshd/rsync/flist.c
2014-12-21 17:56:03 -05:00

3232 lines
86 KiB
C

/*
* Generate and receive file lists.
*
* Copyright (C) 1996 Andrew Tridgell
* Copyright (C) 1996 Paul Mackerras
* Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org>
* Copyright (C) 2002-2014 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 "rounding.h"
#include "inums.h"
#include "io.h"
extern int am_root;
extern int am_server;
extern int am_daemon;
extern int am_sender;
extern int am_generator;
extern int inc_recurse;
extern int always_checksum;
extern int module_id;
extern int ignore_errors;
extern int numeric_ids;
extern int recurse;
extern int use_qsort;
extern int xfer_dirs;
extern int filesfrom_fd;
extern int one_file_system;
extern int copy_dirlinks;
extern int preserve_uid;
extern int preserve_gid;
extern int preserve_acls;
extern int preserve_xattrs;
extern int preserve_links;
extern int preserve_hard_links;
extern int preserve_devices;
extern int preserve_specials;
extern int delete_during;
extern int missing_args;
extern int eol_nulls;
extern int relative_paths;
extern int implied_dirs;
extern int ignore_perishable;
extern int non_perishable_cnt;
extern int prune_empty_dirs;
extern int copy_links;
extern int copy_unsafe_links;
extern int protocol_version;
extern int sanitize_paths;
extern int munge_symlinks;
extern int use_safe_inc_flist;
extern int need_unsorted_flist;
extern int sender_symlink_iconv;
extern int output_needs_newline;
extern int sender_keeps_checksum;
extern int unsort_ndx;
extern uid_t our_uid;
extern struct stats stats;
extern char *filesfrom_host;
extern char *usermap, *groupmap;
extern char curr_dir[MAXPATHLEN];
extern struct chmod_mode_struct *chmod_modes;
extern filter_rule_list filter_list;
extern filter_rule_list daemon_filter_list;
#ifdef ICONV_OPTION
extern int filesfrom_convert;
extern iconv_t ic_send, ic_recv;
#endif
#define PTR_SIZE (sizeof (struct file_struct *))
int io_error;
int checksum_len;
dev_t filesystem_dev; /* used to implement -x */
struct file_list *cur_flist, *first_flist, *dir_flist;
int send_dir_ndx = -1, send_dir_depth = -1;
int flist_cnt = 0; /* how many (non-tmp) file list objects exist */
int file_total = 0; /* total of all active items over all file-lists */
int file_old_total = 0; /* total of active items that will soon be gone */
int flist_eof = 0; /* all the file-lists are now known */
#define NORMAL_NAME 0
#define SLASH_ENDING_NAME 1
#define DOTDIR_NAME 2
#define MISSING_NAME 3
/* Starting from protocol version 26, we always use 64-bit ino_t and dev_t
* internally, even if this platform does not allow files to have 64-bit inums.
* The only exception is if we're on a platform with no 64-bit type at all.
*
* Because we use read_longint() to get these off the wire, if you transfer
* devices or (for protocols < 30) hardlinks with dev or inum > 2**32 to a
* machine with no 64-bit types then you will get an overflow error.
*
* Note that if you transfer devices from a 64-bit-devt machine (say, Solaris)
* to a 32-bit-devt machine (say, Linux-2.2/x86) then the device numbers will
* be truncated. But it's a kind of silly thing to do anyhow. */
/* The tmp_* vars are used as a cache area by make_file() to store data
* that the sender doesn't need to remember in its file list. The data
* will survive just long enough to be used by send_file_entry(). */
static dev_t tmp_rdev;
#ifdef SUPPORT_HARD_LINKS
static int64 tmp_dev = -1, tmp_ino;
#endif
static char tmp_sum[MAX_DIGEST_LEN];
static char empty_sum[MAX_DIGEST_LEN];
static int flist_count_offset; /* for --delete --progress */
static void flist_sort_and_clean(struct file_list *flist, int strip_root);
static void output_flist(struct file_list *flist);
void init_flist(void)
{
if (DEBUG_GTE(FLIST, 4)) {
rprintf(FINFO, "FILE_STRUCT_LEN=%d, EXTRA_LEN=%d\n",
(int)FILE_STRUCT_LEN, (int)EXTRA_LEN);
}
checksum_len = protocol_version < 21 ? 2
: protocol_version < 30 ? MD4_DIGEST_LEN
: MD5_DIGEST_LEN;
}
static int show_filelist_p(void)
{
return INFO_GTE(FLIST, 1) && xfer_dirs && !am_server && !inc_recurse;
}
static void start_filelist_progress(char *kind)
{
rprintf(FCLIENT, "%s ... ", kind);
output_needs_newline = 1;
rflush(FINFO);
}
static void emit_filelist_progress(int count)
{
rprintf(FCLIENT, " %d files...\r", count);
}
static void maybe_emit_filelist_progress(int count)
{
if (INFO_GTE(FLIST, 2) && show_filelist_p() && (count % 100) == 0)
emit_filelist_progress(count);
}
static void finish_filelist_progress(const struct file_list *flist)
{
if (INFO_GTE(FLIST, 2)) {
/* This overwrites the progress line */
rprintf(FINFO, "%d file%sto consider\n",
flist->used, flist->used == 1 ? " " : "s ");
} else {
output_needs_newline = 0;
rprintf(FINFO, "done\n");
}
}
void show_flist_stats(void)
{
/* Nothing yet */
}
/* Stat either a symlink or its referent, depending on the settings of
* copy_links, copy_unsafe_links, etc. Returns -1 on error, 0 on success.
*
* If path is the name of a symlink, then the linkbuf buffer (which must hold
* MAXPATHLEN chars) will be set to the symlink's target string.
*
* The stat structure pointed to by stp will contain information about the
* link or the referent as appropriate, if they exist. */
static int readlink_stat(const char *path, STRUCT_STAT *stp, char *linkbuf)
{
#ifdef SUPPORT_LINKS
if (link_stat(path, stp, copy_dirlinks) < 0)
return -1;
if (S_ISLNK(stp->st_mode)) {
int llen = do_readlink(path, linkbuf, MAXPATHLEN - 1);
if (llen < 0)
return -1;
linkbuf[llen] = '\0';
if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) {
if (INFO_GTE(SYMSAFE, 1)) {
rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n",
path, linkbuf);
}
return x_stat(path, stp, NULL);
}
if (munge_symlinks && am_sender && llen > SYMLINK_PREFIX_LEN
&& strncmp(linkbuf, SYMLINK_PREFIX, SYMLINK_PREFIX_LEN) == 0) {
memmove(linkbuf, linkbuf + SYMLINK_PREFIX_LEN,
llen - SYMLINK_PREFIX_LEN + 1);
}
}
return 0;
#else
return x_stat(path, stp, NULL);
#endif
}
int link_stat(const char *path, STRUCT_STAT *stp, int follow_dirlinks)
{
#ifdef SUPPORT_LINKS
if (copy_links)
return x_stat(path, stp, NULL);
if (x_lstat(path, stp, NULL) < 0)
return -1;
if (follow_dirlinks && S_ISLNK(stp->st_mode)) {
STRUCT_STAT st;
if (x_stat(path, &st, NULL) == 0 && S_ISDIR(st.st_mode))
*stp = st;
}
return 0;
#else
return x_stat(path, stp, NULL);
#endif
}
static inline int is_daemon_excluded(const char *fname, int is_dir)
{
if (daemon_filter_list.head
&& check_filter(&daemon_filter_list, FLOG, fname, is_dir) < 0) {
errno = ENOENT;
return 1;
}
return 0;
}
static inline int path_is_daemon_excluded(char *path, int ignore_filename)
{
if (daemon_filter_list.head) {
char *slash = path;
while ((slash = strchr(slash+1, '/')) != NULL) {
int ret;
*slash = '\0';
ret = check_filter(&daemon_filter_list, FLOG, path, 1);
*slash = '/';
if (ret < 0) {
errno = ENOENT;
return 1;
}
}
if (!ignore_filename
&& check_filter(&daemon_filter_list, FLOG, path, 1) < 0) {
errno = ENOENT;
return 1;
}
}
return 0;
}
/* This function is used to check if a file should be included/excluded
* from the list of files based on its name and type etc. The value of
* filter_level is set to either SERVER_FILTERS or ALL_FILTERS. */
static int is_excluded(const char *fname, int is_dir, int filter_level)
{
#if 0 /* This currently never happens, so avoid a useless compare. */
if (filter_level == NO_FILTERS)
return 0;
#endif
if (is_daemon_excluded(fname, is_dir))
return 1;
if (filter_level != ALL_FILTERS)
return 0;
if (filter_list.head
&& check_filter(&filter_list, FINFO, fname, is_dir) < 0)
return 1;
return 0;
}
static void send_directory(int f, struct file_list *flist,
char *fbuf, int len, int flags);
static const char *pathname, *orig_dir;
static int pathname_len;
/* Make sure flist can hold at least flist->used + extra entries. */
static void flist_expand(struct file_list *flist, int extra)
{
struct file_struct **new_ptr;
if (flist->used + extra <= flist->malloced)
return;
if (flist->malloced < FLIST_START)
flist->malloced = FLIST_START;
else if (flist->malloced >= FLIST_LINEAR)
flist->malloced += FLIST_LINEAR;
else
flist->malloced *= 2;
/* In case count jumped or we are starting the list
* with a known size just set it. */
if (flist->malloced < flist->used + extra)
flist->malloced = flist->used + extra;
new_ptr = realloc_array(flist->files, struct file_struct *,
flist->malloced);
if (DEBUG_GTE(FLIST, 1) && flist->malloced != FLIST_START) {
rprintf(FCLIENT, "[%s] expand file_list pointer array to %s bytes, did%s move\n",
who_am_i(),
big_num(sizeof flist->files[0] * flist->malloced),
(new_ptr == flist->files) ? " not" : "");
}
flist->files = new_ptr;
if (!flist->files)
out_of_memory("flist_expand");
}
static void flist_done_allocating(struct file_list *flist)
{
void *ptr = pool_boundary(flist->file_pool, 8*1024);
if (flist->pool_boundary == ptr)
flist->pool_boundary = NULL; /* list didn't use any pool memory */
else
flist->pool_boundary = ptr;
}
/* Call this with EITHER (1) "file, NULL, 0" to chdir() to the file's
* F_PATHNAME(), or (2) "NULL, dir, dirlen" to chdir() to the supplied dir,
* with dir == NULL taken to be the starting directory, and dirlen < 0
* indicating that strdup(dir) should be called and then the -dirlen length
* value checked to ensure that it is not daemon-excluded. */
int change_pathname(struct file_struct *file, const char *dir, int dirlen)
{
if (dirlen < 0) {
char *cpy = strdup(dir);
if (*cpy != '/')
change_dir(orig_dir, CD_SKIP_CHDIR);
if (path_is_daemon_excluded(cpy, 0))
goto chdir_error;
dir = cpy;
dirlen = -dirlen;
} else {
if (file) {
if (pathname == F_PATHNAME(file))
return 1;
dir = F_PATHNAME(file);
if (dir)
dirlen = strlen(dir);
} else if (pathname == dir)
return 1;
if (dir && *dir != '/')
change_dir(orig_dir, CD_SKIP_CHDIR);
}
pathname = dir;
pathname_len = dirlen;
if (!dir)
dir = orig_dir;
if (!change_dir(dir, CD_NORMAL)) {
chdir_error:
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, errno, "change_dir %s failed", full_fname(dir));
if (dir != orig_dir)
change_dir(orig_dir, CD_NORMAL);
pathname = NULL;
pathname_len = 0;
return 0;
}
return 1;
}
static void send_file_entry(int f, const char *fname, struct file_struct *file,
#ifdef SUPPORT_LINKS
const char *symlink_name, int symlink_len,
#endif
int ndx, int first_ndx)
{
static time_t modtime;
static mode_t mode;
#ifdef SUPPORT_HARD_LINKS
static int64 dev;
#endif
static dev_t rdev;
static uint32 rdev_major;
static uid_t uid;
static gid_t gid;
static const char *user_name, *group_name;
static char lastname[MAXPATHLEN];
#ifdef SUPPORT_HARD_LINKS
int first_hlink_ndx = -1;
#endif
int l1, l2;
int xflags;
/* Initialize starting value of xflags and adjust counts. */
if (S_ISREG(file->mode))
xflags = 0;
else if (S_ISDIR(file->mode)) {
stats.num_dirs++;
if (protocol_version >= 30) {
if (file->flags & FLAG_CONTENT_DIR)
xflags = file->flags & FLAG_TOP_DIR;
else if (file->flags & FLAG_IMPLIED_DIR)
xflags = XMIT_TOP_DIR | XMIT_NO_CONTENT_DIR;
else
xflags = XMIT_NO_CONTENT_DIR;
} else
xflags = file->flags & FLAG_TOP_DIR; /* FLAG_TOP_DIR == XMIT_TOP_DIR */
} else {
if (S_ISLNK(file->mode))
stats.num_symlinks++;
else if (IS_DEVICE(file->mode))
stats.num_devices++;
else if (IS_SPECIAL(file->mode))
stats.num_specials++;
xflags = 0;
}
if (file->mode == mode)
xflags |= XMIT_SAME_MODE;
else
mode = file->mode;
if (preserve_devices && IS_DEVICE(mode)) {
if (protocol_version < 28) {
if (tmp_rdev == rdev)
xflags |= XMIT_SAME_RDEV_pre28;
else
rdev = tmp_rdev;
} else {
rdev = tmp_rdev;
if ((uint32)major(rdev) == rdev_major)
xflags |= XMIT_SAME_RDEV_MAJOR;
else
rdev_major = major(rdev);
if (protocol_version < 30 && (uint32)minor(rdev) <= 0xFFu)
xflags |= XMIT_RDEV_MINOR_8_pre30;
}
} else if (preserve_specials && IS_SPECIAL(mode) && protocol_version < 31) {
/* Special files don't need an rdev number, so just make
* the historical transmission of the value efficient. */
if (protocol_version < 28)
xflags |= XMIT_SAME_RDEV_pre28;
else {
rdev = MAKEDEV(major(rdev), 0);
xflags |= XMIT_SAME_RDEV_MAJOR;
if (protocol_version < 30)
xflags |= XMIT_RDEV_MINOR_8_pre30;
}
} else if (protocol_version < 28)
rdev = MAKEDEV(0, 0);
if (!preserve_uid || ((uid_t)F_OWNER(file) == uid && *lastname))
xflags |= XMIT_SAME_UID;
else {
uid = F_OWNER(file);
if (!numeric_ids) {
user_name = add_uid(uid);
if (inc_recurse && user_name)
xflags |= XMIT_USER_NAME_FOLLOWS;
}
}
if (!preserve_gid || ((gid_t)F_GROUP(file) == gid && *lastname))
xflags |= XMIT_SAME_GID;
else {
gid = F_GROUP(file);
if (!numeric_ids) {
group_name = add_gid(gid);
if (inc_recurse && group_name)
xflags |= XMIT_GROUP_NAME_FOLLOWS;
}
}
if (file->modtime == modtime)
xflags |= XMIT_SAME_TIME;
else
modtime = file->modtime;
if (NSEC_BUMP(file) && protocol_version >= 31)
xflags |= XMIT_MOD_NSEC;
#ifdef SUPPORT_HARD_LINKS
if (tmp_dev != -1) {
if (protocol_version >= 30) {
struct ht_int64_node *np = idev_find(tmp_dev, tmp_ino);
first_hlink_ndx = (int32)(long)np->data - 1;
if (first_hlink_ndx < 0) {
np->data = (void*)(long)(first_ndx + ndx + 1);
xflags |= XMIT_HLINK_FIRST;
}
if (DEBUG_GTE(HLINK, 1)) {
if (first_hlink_ndx >= 0) {
rprintf(FINFO, "[%s] #%d hard-links #%d (%sabbrev)\n",
who_am_i(), first_ndx + ndx, first_hlink_ndx,
first_hlink_ndx >= first_ndx ? "" : "un");
} else if (DEBUG_GTE(HLINK, 3)) {
rprintf(FINFO, "[%s] dev:inode for #%d is %s:%s\n",
who_am_i(), first_ndx + ndx,
big_num(tmp_dev), big_num(tmp_ino));
}
}
} else {
if (tmp_dev == dev) {
if (protocol_version >= 28)
xflags |= XMIT_SAME_DEV_pre30;
} else
dev = tmp_dev;
}
xflags |= XMIT_HLINKED;
}
#endif
for (l1 = 0;
lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255);
l1++) {}
l2 = strlen(fname+l1);
if (l1 > 0)
xflags |= XMIT_SAME_NAME;
if (l2 > 255)
xflags |= XMIT_LONG_NAME;
/* We must make sure we don't send a zero flag byte or the
* other end will terminate the flist transfer. Note that
* the use of XMIT_TOP_DIR on a non-dir has no meaning, so
* it's harmless way to add a bit to the first flag byte. */
if (protocol_version >= 28) {
if (!xflags && !S_ISDIR(mode))
xflags |= XMIT_TOP_DIR;
if ((xflags & 0xFF00) || !xflags) {
xflags |= XMIT_EXTENDED_FLAGS;
write_shortint(f, xflags);
} else
write_byte(f, xflags);
} else {
if (!(xflags & 0xFF))
xflags |= S_ISDIR(mode) ? XMIT_LONG_NAME : XMIT_TOP_DIR;
write_byte(f, xflags);
}
if (xflags & XMIT_SAME_NAME)
write_byte(f, l1);
if (xflags & XMIT_LONG_NAME)
write_varint30(f, l2);
else
write_byte(f, l2);
write_buf(f, fname + l1, l2);
#ifdef SUPPORT_HARD_LINKS
if (first_hlink_ndx >= 0) {
write_varint(f, first_hlink_ndx);
if (first_hlink_ndx >= first_ndx)
goto the_end;
}
#endif
write_varlong30(f, F_LENGTH(file), 3);
if (!(xflags & XMIT_SAME_TIME)) {
if (protocol_version >= 30)
write_varlong(f, modtime, 4);
else
write_int(f, modtime);
}
if (xflags & XMIT_MOD_NSEC)
write_varint(f, F_MOD_NSEC(file));
if (!(xflags & XMIT_SAME_MODE))
write_int(f, to_wire_mode(mode));
if (preserve_uid && !(xflags & XMIT_SAME_UID)) {
if (protocol_version < 30)
write_int(f, uid);
else {
write_varint(f, uid);
if (xflags & XMIT_USER_NAME_FOLLOWS) {
int len = strlen(user_name);
write_byte(f, len);
write_buf(f, user_name, len);
}
}
}
if (preserve_gid && !(xflags & XMIT_SAME_GID)) {
if (protocol_version < 30)
write_int(f, gid);
else {
write_varint(f, gid);
if (xflags & XMIT_GROUP_NAME_FOLLOWS) {
int len = strlen(group_name);
write_byte(f, len);
write_buf(f, group_name, len);
}
}
}
if ((preserve_devices && IS_DEVICE(mode))
|| (preserve_specials && IS_SPECIAL(mode) && protocol_version < 31)) {
if (protocol_version < 28) {
if (!(xflags & XMIT_SAME_RDEV_pre28))
write_int(f, (int)rdev);
} else {
if (!(xflags & XMIT_SAME_RDEV_MAJOR))
write_varint30(f, major(rdev));
if (protocol_version >= 30)
write_varint(f, minor(rdev));
else if (xflags & XMIT_RDEV_MINOR_8_pre30)
write_byte(f, minor(rdev));
else
write_int(f, minor(rdev));
}
}
#ifdef SUPPORT_LINKS
if (symlink_len) {
write_varint30(f, symlink_len);
write_buf(f, symlink_name, symlink_len);
}
#endif
#ifdef SUPPORT_HARD_LINKS
if (tmp_dev != -1 && protocol_version < 30) {
/* Older protocols expect the dev number to be transmitted
* 1-incremented so that it is never zero. */
if (protocol_version < 26) {
/* 32-bit dev_t and ino_t */
write_int(f, (int32)(dev+1));
write_int(f, (int32)tmp_ino);
} else {
/* 64-bit dev_t and ino_t */
if (!(xflags & XMIT_SAME_DEV_pre30))
write_longint(f, dev+1);
write_longint(f, tmp_ino);
}
}
#endif
if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
const char *sum;
if (S_ISREG(mode))
sum = tmp_sum;
else {
/* Prior to 28, we sent a useless set of nulls. */
sum = empty_sum;
}
write_buf(f, sum, checksum_len);
}
#ifdef SUPPORT_HARD_LINKS
the_end:
#endif
strlcpy(lastname, fname, MAXPATHLEN);
if (S_ISREG(mode) || S_ISLNK(mode))
stats.total_size += F_LENGTH(file);
}
static struct file_struct *recv_file_entry(int f, struct file_list *flist, int xflags)
{
static int64 modtime;
static mode_t mode;
#ifdef SUPPORT_HARD_LINKS
static int64 dev;
#endif
static dev_t rdev;
static uint32 rdev_major;
static uid_t uid;
static gid_t gid;
static uint16 gid_flags;
static char lastname[MAXPATHLEN], *lastdir;
static int lastdir_depth, lastdir_len = -1;
static unsigned int del_hier_name_len = 0;
static int in_del_hier = 0;
char thisname[MAXPATHLEN];
unsigned int l1 = 0, l2 = 0;
int alloc_len, basename_len, linkname_len;
int extra_len = file_extra_cnt * EXTRA_LEN;
int first_hlink_ndx = -1;
int64 file_length;
uint32 modtime_nsec;
const char *basename;
struct file_struct *file;
alloc_pool_t *pool;
char *bp;
if (xflags & XMIT_SAME_NAME)
l1 = read_byte(f);
if (xflags & XMIT_LONG_NAME)
l2 = read_varint30(f);
else
l2 = read_byte(f);
if (l2 >= MAXPATHLEN - l1) {
rprintf(FERROR,
"overflow: xflags=0x%x l1=%d l2=%d lastname=%s [%s]\n",
xflags, l1, l2, lastname, who_am_i());
overflow_exit("recv_file_entry");
}
strlcpy(thisname, lastname, l1 + 1);
read_sbuf(f, &thisname[l1], l2);
thisname[l1 + l2] = 0;
/* Abuse basename_len for a moment... */
basename_len = strlcpy(lastname, thisname, MAXPATHLEN);
#ifdef ICONV_OPTION
if (ic_recv != (iconv_t)-1) {
xbuf outbuf, inbuf;
INIT_CONST_XBUF(outbuf, thisname);
INIT_XBUF(inbuf, lastname, basename_len, (size_t)-1);
if (iconvbufs(ic_recv, &inbuf, &outbuf, ICB_INIT) < 0) {
io_error |= IOERR_GENERAL;
rprintf(FERROR_UTF8,
"[%s] cannot convert filename: %s (%s)\n",
who_am_i(), lastname, strerror(errno));
outbuf.len = 0;
}
thisname[outbuf.len] = '\0';
}
#endif
if (*thisname
&& (clean_fname(thisname, CFN_REFUSE_DOT_DOT_DIRS) < 0 || (!relative_paths && *thisname == '/'))) {
rprintf(FERROR, "ABORTING due to unsafe pathname from sender: %s\n", thisname);
exit_cleanup(RERR_PROTOCOL);
}
if (sanitize_paths)
sanitize_path(thisname, thisname, "", 0, SP_DEFAULT);
if ((basename = strrchr(thisname, '/')) != NULL) {
int len = basename++ - thisname;
if (len != lastdir_len || memcmp(thisname, lastdir, len) != 0) {
lastdir = new_array(char, len + 1);
memcpy(lastdir, thisname, len);
lastdir[len] = '\0';
lastdir_len = len;
lastdir_depth = count_dir_elements(lastdir);
}
} else
basename = thisname;
basename_len = strlen(basename) + 1; /* count the '\0' */
#ifdef SUPPORT_HARD_LINKS
if (protocol_version >= 30
&& BITS_SETnUNSET(xflags, XMIT_HLINKED, XMIT_HLINK_FIRST)) {
first_hlink_ndx = read_varint(f);
if (first_hlink_ndx < 0 || first_hlink_ndx >= flist->ndx_start + flist->used) {
rprintf(FERROR,
"hard-link reference out of range: %d (%d)\n",
first_hlink_ndx, flist->ndx_start + flist->used);
exit_cleanup(RERR_PROTOCOL);
}
if (DEBUG_GTE(HLINK, 1)) {
rprintf(FINFO, "[%s] #%d hard-links #%d (%sabbrev)\n",
who_am_i(), flist->used+flist->ndx_start, first_hlink_ndx,
first_hlink_ndx >= flist->ndx_start ? "" : "un");
}
if (first_hlink_ndx >= flist->ndx_start) {
struct file_struct *first = flist->files[first_hlink_ndx - flist->ndx_start];
file_length = F_LENGTH(first);
modtime = first->modtime;
modtime_nsec = F_MOD_NSEC(first);
mode = first->mode;
if (preserve_uid)
uid = F_OWNER(first);
if (preserve_gid)
gid = F_GROUP(first);
if (preserve_devices && IS_DEVICE(mode)) {
uint32 *devp = F_RDEV_P(first);
rdev = MAKEDEV(DEV_MAJOR(devp), DEV_MINOR(devp));
extra_len += DEV_EXTRA_CNT * EXTRA_LEN;
}
if (preserve_links && S_ISLNK(mode))
linkname_len = strlen(F_SYMLINK(first)) + 1;
else
linkname_len = 0;
goto create_object;
}
}
#endif
file_length = read_varlong30(f, 3);
if (!(xflags & XMIT_SAME_TIME)) {
if (protocol_version >= 30) {
modtime = read_varlong(f, 4);
#if SIZEOF_TIME_T < SIZEOF_INT64
if (!am_generator && (int64)(time_t)modtime != modtime) {
rprintf(FERROR_XFER,
"Time value of %s truncated on receiver.\n",
lastname);
}
#endif
} else
modtime = read_int(f);
}
if (xflags & XMIT_MOD_NSEC)
modtime_nsec = read_varint(f);
else
modtime_nsec = 0;
if (!(xflags & XMIT_SAME_MODE))
mode = from_wire_mode(read_int(f));
if (chmod_modes && !S_ISLNK(mode) && mode)
mode = tweak_mode(mode, chmod_modes);
if (preserve_uid && !(xflags & XMIT_SAME_UID)) {
if (protocol_version < 30)
uid = (uid_t)read_int(f);
else {
uid = (uid_t)read_varint(f);
if (xflags & XMIT_USER_NAME_FOLLOWS)
uid = recv_user_name(f, uid);
else if (inc_recurse && am_root && (!numeric_ids || usermap))
uid = match_uid(uid);
}
}
if (preserve_gid && !(xflags & XMIT_SAME_GID)) {
if (protocol_version < 30)
gid = (gid_t)read_int(f);
else {
gid = (gid_t)read_varint(f);
gid_flags = 0;
if (xflags & XMIT_GROUP_NAME_FOLLOWS)
gid = recv_group_name(f, gid, &gid_flags);
else if (inc_recurse && (!am_root || !numeric_ids || groupmap))
gid = match_gid(gid, &gid_flags);
}
}
if ((preserve_devices && IS_DEVICE(mode))
|| (preserve_specials && IS_SPECIAL(mode) && protocol_version < 31)) {
if (protocol_version < 28) {
if (!(xflags & XMIT_SAME_RDEV_pre28))
rdev = (dev_t)read_int(f);
} else {
uint32 rdev_minor;
if (!(xflags & XMIT_SAME_RDEV_MAJOR))
rdev_major = read_varint30(f);
if (protocol_version >= 30)
rdev_minor = read_varint(f);
else if (xflags & XMIT_RDEV_MINOR_8_pre30)
rdev_minor = read_byte(f);
else
rdev_minor = read_int(f);
rdev = MAKEDEV(rdev_major, rdev_minor);
}
if (IS_DEVICE(mode))
extra_len += DEV_EXTRA_CNT * EXTRA_LEN;
file_length = 0;
} else if (protocol_version < 28)
rdev = MAKEDEV(0, 0);
#ifdef SUPPORT_LINKS
if (preserve_links && S_ISLNK(mode)) {
linkname_len = read_varint30(f) + 1; /* count the '\0' */
if (linkname_len <= 0 || linkname_len > MAXPATHLEN) {
rprintf(FERROR, "overflow: linkname_len=%d\n",
linkname_len - 1);
overflow_exit("recv_file_entry");
}
#ifdef ICONV_OPTION
/* We don't know how much extra room we need to convert
* the as-yet-unread symlink data, so let's hope that a
* double-size buffer is plenty. */
if (sender_symlink_iconv)
linkname_len *= 2;
#endif
if (munge_symlinks)
linkname_len += SYMLINK_PREFIX_LEN;
}
else
#endif
linkname_len = 0;
#ifdef SUPPORT_HARD_LINKS
create_object:
if (preserve_hard_links) {
if (protocol_version < 28 && S_ISREG(mode))
xflags |= XMIT_HLINKED;
if (xflags & XMIT_HLINKED)
extra_len += (inc_recurse+1) * EXTRA_LEN;
}
#endif
#ifdef SUPPORT_ACLS
/* Directories need an extra int32 for the default ACL. */
if (preserve_acls && S_ISDIR(mode))
extra_len += EXTRA_LEN;
#endif
if (always_checksum && S_ISREG(mode))
extra_len += SUM_EXTRA_CNT * EXTRA_LEN;
#if SIZEOF_INT64 >= 8
if (file_length > 0xFFFFFFFFu && S_ISREG(mode))
extra_len += EXTRA_LEN;
#endif
#ifdef HAVE_UTIMENSAT
if (modtime_nsec)
extra_len += EXTRA_LEN;
#endif
if (file_length < 0) {
rprintf(FERROR, "Offset underflow: file-length is negative\n");
exit_cleanup(RERR_UNSUPPORTED);
}
if (inc_recurse && S_ISDIR(mode)) {
if (one_file_system) {
/* Room to save the dir's device for -x */
extra_len += DEV_EXTRA_CNT * EXTRA_LEN;
}
pool = dir_flist->file_pool;
} else
pool = flist->file_pool;
#if EXTRA_ROUNDING > 0
if (extra_len & (EXTRA_ROUNDING * EXTRA_LEN))
extra_len = (extra_len | (EXTRA_ROUNDING * EXTRA_LEN)) + EXTRA_LEN;
#endif
alloc_len = FILE_STRUCT_LEN + extra_len + basename_len
+ linkname_len;
bp = pool_alloc(pool, alloc_len, "recv_file_entry");
memset(bp, 0, extra_len + FILE_STRUCT_LEN);
bp += extra_len;
file = (struct file_struct *)bp;
bp += FILE_STRUCT_LEN;
memcpy(bp, basename, basename_len);
#ifdef SUPPORT_HARD_LINKS
if (xflags & XMIT_HLINKED
#ifndef CAN_HARDLINK_SYMLINK
&& !S_ISLNK(mode)
#endif
#ifndef CAN_HARDLINK_SPECIAL
&& !IS_SPECIAL(mode) && !IS_DEVICE(mode)
#endif
)
file->flags |= FLAG_HLINKED;
#endif
file->modtime = (time_t)modtime;
#ifdef HAVE_UTIMENSAT
if (modtime_nsec) {
file->flags |= FLAG_MOD_NSEC;
OPT_EXTRA(file, 0)->unum = modtime_nsec;
}
#endif
file->len32 = (uint32)file_length;
#if SIZEOF_INT64 >= 8
if (file_length > 0xFFFFFFFFu && S_ISREG(mode)) {
#if SIZEOF_CAPITAL_OFF_T < 8
rprintf(FERROR, "Offset overflow: attempted 64-bit file-length\n");
exit_cleanup(RERR_UNSUPPORTED);
#else
file->flags |= FLAG_LENGTH64;
OPT_EXTRA(file, NSEC_BUMP(file))->unum = (uint32)(file_length >> 32);
#endif
}
#endif
file->mode = mode;
if (preserve_uid)
F_OWNER(file) = uid;
if (preserve_gid) {
F_GROUP(file) = gid;
file->flags |= gid_flags;
}
if (unsort_ndx)
F_NDX(file) = flist->used + flist->ndx_start;
if (basename != thisname) {
file->dirname = lastdir;
F_DEPTH(file) = lastdir_depth + 1;
} else
F_DEPTH(file) = 1;
if (S_ISDIR(mode)) {
if (basename_len == 1+1 && *basename == '.') /* +1 for '\0' */
F_DEPTH(file)--;
if (protocol_version >= 30) {
if (!(xflags & XMIT_NO_CONTENT_DIR)) {
if (xflags & XMIT_TOP_DIR)
file->flags |= FLAG_TOP_DIR;
file->flags |= FLAG_CONTENT_DIR;
} else if (xflags & XMIT_TOP_DIR)
file->flags |= FLAG_IMPLIED_DIR;
} else if (xflags & XMIT_TOP_DIR) {
in_del_hier = recurse;
del_hier_name_len = F_DEPTH(file) == 0 ? 0 : l1 + l2;
if (relative_paths && del_hier_name_len > 2
&& lastname[del_hier_name_len-1] == '.'
&& lastname[del_hier_name_len-2] == '/')
del_hier_name_len -= 2;
file->flags |= FLAG_TOP_DIR | FLAG_CONTENT_DIR;
} else if (in_del_hier) {
if (!relative_paths || !del_hier_name_len
|| (l1 >= del_hier_name_len
&& lastname[del_hier_name_len] == '/'))
file->flags |= FLAG_CONTENT_DIR;
else
in_del_hier = 0;
}
}
if (preserve_devices && IS_DEVICE(mode)) {
uint32 *devp = F_RDEV_P(file);
DEV_MAJOR(devp) = major(rdev);
DEV_MINOR(devp) = minor(rdev);
}
#ifdef SUPPORT_LINKS
if (linkname_len) {
bp += basename_len;
if (first_hlink_ndx >= flist->ndx_start) {
struct file_struct *first = flist->files[first_hlink_ndx - flist->ndx_start];
memcpy(bp, F_SYMLINK(first), linkname_len);
} else {
if (munge_symlinks) {
strlcpy(bp, SYMLINK_PREFIX, linkname_len);
bp += SYMLINK_PREFIX_LEN;
linkname_len -= SYMLINK_PREFIX_LEN;
}
#ifdef ICONV_OPTION
if (sender_symlink_iconv) {
xbuf outbuf, inbuf;
alloc_len = linkname_len;
linkname_len /= 2;
/* Read the symlink data into the end of our double-sized
* buffer and then convert it into the right spot. */
INIT_XBUF(inbuf, bp + alloc_len - linkname_len,
linkname_len - 1, (size_t)-1);
read_sbuf(f, inbuf.buf, inbuf.len);
INIT_XBUF(outbuf, bp, 0, alloc_len);
if (iconvbufs(ic_recv, &inbuf, &outbuf, ICB_INIT) < 0) {
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER,
"[%s] cannot convert symlink data for: %s (%s)\n",
who_am_i(), full_fname(thisname), strerror(errno));
bp = (char*)file->basename;
*bp++ = '\0';
outbuf.len = 0;
}
bp[outbuf.len] = '\0';
} else
#endif
read_sbuf(f, bp, linkname_len - 1);
if (sanitize_paths && !munge_symlinks && *bp)
sanitize_path(bp, bp, "", lastdir_depth, SP_DEFAULT);
}
}
#endif
#ifdef SUPPORT_HARD_LINKS
if (preserve_hard_links && xflags & XMIT_HLINKED) {
if (protocol_version >= 30) {
if (xflags & XMIT_HLINK_FIRST) {
F_HL_GNUM(file) = flist->ndx_start + flist->used;
} else
F_HL_GNUM(file) = first_hlink_ndx;
} else {
static int32 cnt = 0;
struct ht_int64_node *np;
int64 ino;
int32 ndx;
if (protocol_version < 26) {
dev = read_int(f);
ino = read_int(f);
} else {
if (!(xflags & XMIT_SAME_DEV_pre30))
dev = read_longint(f);
ino = read_longint(f);
}
np = idev_find(dev, ino);
ndx = (int32)(long)np->data - 1;
if (ndx < 0) {
ndx = cnt++;
np->data = (void*)(long)cnt;
}
F_HL_GNUM(file) = ndx;
}
}
#endif
if (always_checksum && (S_ISREG(mode) || protocol_version < 28)) {
if (S_ISREG(mode))
bp = F_SUM(file);
else {
/* Prior to 28, we get a useless set of nulls. */
bp = tmp_sum;
}
if (first_hlink_ndx >= flist->ndx_start) {
struct file_struct *first = flist->files[first_hlink_ndx - flist->ndx_start];
memcpy(bp, F_SUM(first), checksum_len);
} else
read_buf(f, bp, checksum_len);
}
#ifdef SUPPORT_ACLS
if (preserve_acls && !S_ISLNK(mode))
receive_acl(f, file);
#endif
#ifdef SUPPORT_XATTRS
if (preserve_xattrs)
receive_xattr(f, file);
#endif
if (S_ISREG(mode) || S_ISLNK(mode))
stats.total_size += file_length;
return file;
}
/* Create a file_struct for a named file by reading its stat() information
* and performing extensive checks against global options.
*
* Returns a pointer to the new file struct, or NULL if there was an error
* or this file should be excluded.
*
* Note: Any error (here or in send_file_name) that results in the omission of
* an existent source file from the file list should set
* "io_error |= IOERR_GENERAL" to avoid deletion of the file from the
* destination if --delete is on. */
struct file_struct *make_file(const char *fname, struct file_list *flist,
STRUCT_STAT *stp, int flags, int filter_level)
{
static char *lastdir;
static int lastdir_len = -1;
struct file_struct *file;
char thisname[MAXPATHLEN];
char linkname[MAXPATHLEN];
int alloc_len, basename_len, linkname_len;
int extra_len = file_extra_cnt * EXTRA_LEN;
const char *basename;
alloc_pool_t *pool;
STRUCT_STAT st;
char *bp;
if (strlcpy(thisname, fname, sizeof thisname) >= sizeof thisname) {
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER, "skipping overly long name: %s\n", fname);
return NULL;
}
clean_fname(thisname, 0);
if (sanitize_paths)
sanitize_path(thisname, thisname, "", 0, SP_DEFAULT);
if (stp && (S_ISDIR(stp->st_mode) || IS_MISSING_FILE(*stp))) {
/* This is needed to handle a "symlink/." with a --relative
* dir, or a request to delete a specific file. */
st = *stp;
*linkname = '\0'; /* make IBM code checker happy */
} else if (readlink_stat(thisname, &st, linkname) != 0) {
int save_errno = errno;
/* See if file is excluded before reporting an error. */
if (filter_level != NO_FILTERS
&& (is_excluded(thisname, 0, filter_level)
|| is_excluded(thisname, 1, filter_level))) {
if (ignore_perishable && save_errno != ENOENT)
non_perishable_cnt++;
return NULL;
}
if (save_errno == ENOENT) {
#ifdef SUPPORT_LINKS
/* When our options tell us to follow a symlink that
* points nowhere, tell the user about the symlink
* instead of giving a "vanished" message. We only
* dereference a symlink if one of the --copy*links
* options was specified, so there's no need for the
* extra lstat() if one of these options isn't on. */
if ((copy_links || copy_unsafe_links || copy_dirlinks)
&& x_lstat(thisname, &st, NULL) == 0
&& S_ISLNK(st.st_mode)) {
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER, "symlink has no referent: %s\n",
full_fname(thisname));
} else
#endif
{
enum logcode c = am_daemon && protocol_version < 28
? FERROR : FWARNING;
io_error |= IOERR_VANISHED;
rprintf(c, "file has vanished: %s\n",
full_fname(thisname));
}
} else {
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, save_errno, "readlink_stat(%s) failed",
full_fname(thisname));
}
return NULL;
} else if (IS_MISSING_FILE(st)) {
io_error |= IOERR_GENERAL;
rprintf(FINFO, "skipping file with bogus (zero) st_mode: %s\n",
full_fname(thisname));
return NULL;
}
if (filter_level == NO_FILTERS)
goto skip_filters;
if (S_ISDIR(st.st_mode)) {
if (!xfer_dirs) {
rprintf(FINFO, "skipping directory %s\n", thisname);
return NULL;
}
/* -x only affects dirs because we need to avoid recursing
* into a mount-point directory, not to avoid copying a
* symlinked file if -L (or similar) was specified. */
if (one_file_system && st.st_dev != filesystem_dev
&& BITS_SETnUNSET(flags, FLAG_CONTENT_DIR, FLAG_TOP_DIR)) {
if (one_file_system > 1) {
if (INFO_GTE(MOUNT, 1)) {
rprintf(FINFO,
"[%s] skipping mount-point dir %s\n",
who_am_i(), thisname);
}
return NULL;
}
flags |= FLAG_MOUNT_DIR;
flags &= ~FLAG_CONTENT_DIR;
}
} else
flags &= ~FLAG_CONTENT_DIR;
if (is_excluded(thisname, S_ISDIR(st.st_mode) != 0, filter_level)) {
if (ignore_perishable)
non_perishable_cnt++;
return NULL;
}
if (lp_ignore_nonreadable(module_id)) {
#ifdef SUPPORT_LINKS
if (!S_ISLNK(st.st_mode))
#endif
if (access(thisname, R_OK) != 0)
return NULL;
}
skip_filters:
/* Only divert a directory in the main transfer. */
if (flist) {
if (flist->prev && S_ISDIR(st.st_mode)
&& flags & FLAG_DIVERT_DIRS) {
/* Room for parent/sibling/next-child info. */
extra_len += DIRNODE_EXTRA_CNT * EXTRA_LEN;
if (relative_paths)
extra_len += PTR_EXTRA_CNT * EXTRA_LEN;
pool = dir_flist->file_pool;
} else
pool = flist->file_pool;
} else {
#ifdef SUPPORT_ACLS
/* Directories need an extra int32 for the default ACL. */
if (preserve_acls && S_ISDIR(st.st_mode))
extra_len += EXTRA_LEN;
#endif
pool = NULL;
}
if (DEBUG_GTE(FLIST, 2)) {
rprintf(FINFO, "[%s] make_file(%s,*,%d)\n",
who_am_i(), thisname, filter_level);
}
if ((basename = strrchr(thisname, '/')) != NULL) {
int len = basename++ - thisname;
if (len != lastdir_len || memcmp(thisname, lastdir, len) != 0) {
lastdir = new_array(char, len + 1);
memcpy(lastdir, thisname, len);
lastdir[len] = '\0';
lastdir_len = len;
}
} else
basename = thisname;
basename_len = strlen(basename) + 1; /* count the '\0' */
#ifdef SUPPORT_LINKS
linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0;
#else
linkname_len = 0;
#endif
#ifdef ST_MTIME_NSEC
if (st.ST_MTIME_NSEC && protocol_version >= 31)
extra_len += EXTRA_LEN;
#endif
#if SIZEOF_CAPITAL_OFF_T >= 8
if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode))
extra_len += EXTRA_LEN;
#endif
if (always_checksum && am_sender && S_ISREG(st.st_mode)) {
file_checksum(thisname, &st, tmp_sum);
if (sender_keeps_checksum)
extra_len += SUM_EXTRA_CNT * EXTRA_LEN;
}
#if EXTRA_ROUNDING > 0
if (extra_len & (EXTRA_ROUNDING * EXTRA_LEN))
extra_len = (extra_len | (EXTRA_ROUNDING * EXTRA_LEN)) + EXTRA_LEN;
#endif
alloc_len = FILE_STRUCT_LEN + extra_len + basename_len
+ linkname_len;
if (pool)
bp = pool_alloc(pool, alloc_len, "make_file");
else {
if (!(bp = new_array(char, alloc_len)))
out_of_memory("make_file");
}
memset(bp, 0, extra_len + FILE_STRUCT_LEN);
bp += extra_len;
file = (struct file_struct *)bp;
bp += FILE_STRUCT_LEN;
memcpy(bp, basename, basename_len);
#ifdef SUPPORT_HARD_LINKS
if (preserve_hard_links && flist && flist->prev) {
if (protocol_version >= 28
? (!S_ISDIR(st.st_mode) && st.st_nlink > 1)
: S_ISREG(st.st_mode)) {
tmp_dev = (int64)st.st_dev;
tmp_ino = (int64)st.st_ino;
} else
tmp_dev = -1;
}
#endif
#ifdef HAVE_STRUCT_STAT_ST_RDEV
if (IS_DEVICE(st.st_mode)) {
tmp_rdev = st.st_rdev;
st.st_size = 0;
} else if (IS_SPECIAL(st.st_mode))
st.st_size = 0;
#endif
file->flags = flags;
file->modtime = st.st_mtime;
#ifdef ST_MTIME_NSEC
if (st.ST_MTIME_NSEC && protocol_version >= 31) {
file->flags |= FLAG_MOD_NSEC;
OPT_EXTRA(file, 0)->unum = st.ST_MTIME_NSEC;
}
#endif
file->len32 = (uint32)st.st_size;
#if SIZEOF_CAPITAL_OFF_T >= 8
if (st.st_size > 0xFFFFFFFFu && S_ISREG(st.st_mode)) {
file->flags |= FLAG_LENGTH64;
OPT_EXTRA(file, NSEC_BUMP(file))->unum = (uint32)(st.st_size >> 32);
}
#endif
file->mode = st.st_mode;
if (preserve_uid)
F_OWNER(file) = st.st_uid;
if (preserve_gid)
F_GROUP(file) = st.st_gid;
if (am_generator && st.st_uid == our_uid)
file->flags |= FLAG_OWNED_BY_US;
if (basename != thisname)
file->dirname = lastdir;
#ifdef SUPPORT_LINKS
if (linkname_len)
memcpy(bp + basename_len, linkname, linkname_len);
#endif
if (am_sender)
F_PATHNAME(file) = pathname;
else if (!pool)
F_DEPTH(file) = extra_len / EXTRA_LEN;
if (basename_len == 0+1) {
if (!pool)
unmake_file(file);
return NULL;
}
if (sender_keeps_checksum && S_ISREG(st.st_mode))
memcpy(F_SUM(file), tmp_sum, checksum_len);
if (unsort_ndx)
F_NDX(file) = stats.num_dirs;
return file;
}
/* Only called for temporary file_struct entries created by make_file(). */
void unmake_file(struct file_struct *file)
{
free(REQ_EXTRA(file, F_DEPTH(file)));
}
static struct file_struct *send_file_name(int f, struct file_list *flist,
const char *fname, STRUCT_STAT *stp,
int flags, int filter_level)
{
struct file_struct *file;
file = make_file(fname, flist, stp, flags, filter_level);
if (!file)
return NULL;
if (chmod_modes && !S_ISLNK(file->mode) && file->mode)
file->mode = tweak_mode(file->mode, chmod_modes);
if (f >= 0) {
char fbuf[MAXPATHLEN];
#ifdef SUPPORT_LINKS
const char *symlink_name;
int symlink_len;
#ifdef ICONV_OPTION
char symlink_buf[MAXPATHLEN];
#endif
#endif
#if defined SUPPORT_ACLS || defined SUPPORT_XATTRS
stat_x sx;
init_stat_x(&sx);
#endif
#ifdef SUPPORT_LINKS
if (preserve_links && S_ISLNK(file->mode)) {
symlink_name = F_SYMLINK(file);
symlink_len = strlen(symlink_name);
if (symlink_len == 0) {
io_error |= IOERR_GENERAL;
f_name(file, fbuf);
rprintf(FERROR_XFER,
"skipping symlink with 0-length value: %s\n",
full_fname(fbuf));
return NULL;
}
} else {
symlink_name = NULL;
symlink_len = 0;
}
#endif
#ifdef ICONV_OPTION
if (ic_send != (iconv_t)-1) {
xbuf outbuf, inbuf;
INIT_CONST_XBUF(outbuf, fbuf);
if (file->dirname) {
INIT_XBUF_STRLEN(inbuf, (char*)file->dirname);
outbuf.size -= 2; /* Reserve room for '/' & 1 more char. */
if (iconvbufs(ic_send, &inbuf, &outbuf, ICB_INIT) < 0)
goto convert_error;
outbuf.size += 2;
fbuf[outbuf.len++] = '/';
}
INIT_XBUF_STRLEN(inbuf, (char*)file->basename);
if (iconvbufs(ic_send, &inbuf, &outbuf, ICB_INIT) < 0) {
convert_error:
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER,
"[%s] cannot convert filename: %s (%s)\n",
who_am_i(), f_name(file, fbuf), strerror(errno));
return NULL;
}
fbuf[outbuf.len] = '\0';
#ifdef SUPPORT_LINKS
if (symlink_len && sender_symlink_iconv) {
INIT_XBUF(inbuf, (char*)symlink_name, symlink_len, (size_t)-1);
INIT_CONST_XBUF(outbuf, symlink_buf);
if (iconvbufs(ic_send, &inbuf, &outbuf, ICB_INIT) < 0) {
io_error |= IOERR_GENERAL;
f_name(file, fbuf);
rprintf(FERROR_XFER,
"[%s] cannot convert symlink data for: %s (%s)\n",
who_am_i(), full_fname(fbuf), strerror(errno));
return NULL;
}
symlink_buf[outbuf.len] = '\0';
symlink_name = symlink_buf;
symlink_len = outbuf.len;
}
#endif
} else
#endif
f_name(file, fbuf);
#ifdef SUPPORT_ACLS
if (preserve_acls && !S_ISLNK(file->mode)) {
sx.st.st_mode = file->mode;
if (get_acl(fname, &sx) < 0) {
io_error |= IOERR_GENERAL;
return NULL;
}
}
#endif
#ifdef SUPPORT_XATTRS
if (preserve_xattrs) {
sx.st.st_mode = file->mode;
if (get_xattr(fname, &sx) < 0) {
io_error |= IOERR_GENERAL;
return NULL;
}
}
#endif
send_file_entry(f, fbuf, file,
#ifdef SUPPORT_LINKS
symlink_name, symlink_len,
#endif
flist->used, flist->ndx_start);
#ifdef SUPPORT_ACLS
if (preserve_acls && !S_ISLNK(file->mode)) {
send_acl(f, &sx);
free_acl(&sx);
}
#endif
#ifdef SUPPORT_XATTRS
if (preserve_xattrs) {
F_XATTR(file) = send_xattr(f, &sx);
free_xattr(&sx);
}
#endif
}
maybe_emit_filelist_progress(flist->used + flist_count_offset);
flist_expand(flist, 1);
flist->files[flist->used++] = file;
return file;
}
static void send_if_directory(int f, struct file_list *flist,
struct file_struct *file,
char *fbuf, unsigned int ol,
int flags)
{
char is_dot_dir = fbuf[ol-1] == '.' && (ol == 1 || fbuf[ol-2] == '/');
if (S_ISDIR(file->mode)
&& !(file->flags & FLAG_MOUNT_DIR) && f_name(file, fbuf)) {
void *save_filters;
unsigned int len = strlen(fbuf);
if (len > 1 && fbuf[len-1] == '/')
fbuf[--len] = '\0';
save_filters = push_local_filters(fbuf, len);
send_directory(f, flist, fbuf, len, flags);
pop_local_filters(save_filters);
fbuf[ol] = '\0';
if (is_dot_dir)
fbuf[ol-1] = '.';
}
}
static int file_compare(const void *file1, const void *file2)
{
return f_name_cmp(*(struct file_struct **)file1,
*(struct file_struct **)file2);
}
/* The guts of a merge-sort algorithm. This was derived from the glibc
* version, but I (Wayne) changed the merge code to do less copying and
* to require only half the amount of temporary memory. */
static void fsort_tmp(struct file_struct **fp, size_t num,
struct file_struct **tmp)
{
struct file_struct **f1, **f2, **t;
size_t n1, n2;
n1 = num / 2;
n2 = num - n1;
f1 = fp;
f2 = fp + n1;
if (n1 > 1)
fsort_tmp(f1, n1, tmp);
if (n2 > 1)
fsort_tmp(f2, n2, tmp);
while (f_name_cmp(*f1, *f2) <= 0) {
if (!--n1)
return;
f1++;
}
t = tmp;
memcpy(t, f1, n1 * PTR_SIZE);
*f1++ = *f2++, n2--;
while (n1 > 0 && n2 > 0) {
if (f_name_cmp(*t, *f2) <= 0)
*f1++ = *t++, n1--;
else
*f1++ = *f2++, n2--;
}
if (n1 > 0)
memcpy(f1, t, n1 * PTR_SIZE);
}
/* This file-struct sorting routine makes sure that any identical names in
* the file list stay in the same order as they were in the original list.
* This is particularly vital in inc_recurse mode where we expect a sort
* on the flist to match the exact order of a sort on the dir_flist. */
static void fsort(struct file_struct **fp, size_t num)
{
if (num <= 1)
return;
if (use_qsort)
qsort(fp, num, PTR_SIZE, file_compare);
else {
struct file_struct **tmp = new_array(struct file_struct *,
(num+1) / 2);
fsort_tmp(fp, num, tmp);
free(tmp);
}
}
/* We take an entire set of sibling dirs from the sorted flist and link them
* into the tree, setting the appropriate parent/child/sibling pointers. */
static void add_dirs_to_tree(int parent_ndx, struct file_list *from_flist,
int dir_cnt)
{
int i;
int32 *dp = NULL;
int32 *parent_dp = parent_ndx < 0 ? NULL
: F_DIR_NODE_P(dir_flist->sorted[parent_ndx]);
flist_expand(dir_flist, dir_cnt);
dir_flist->sorted = dir_flist->files;
for (i = 0; dir_cnt; i++) {
struct file_struct *file = from_flist->sorted[i];
if (!S_ISDIR(file->mode))
continue;
dir_flist->files[dir_flist->used++] = file;
dir_cnt--;
if (file->basename[0] == '.' && file->basename[1] == '\0')
continue;
if (dp)
DIR_NEXT_SIBLING(dp) = dir_flist->used - 1;
else if (parent_dp)
DIR_FIRST_CHILD(parent_dp) = dir_flist->used - 1;
else
send_dir_ndx = dir_flist->used - 1;
dp = F_DIR_NODE_P(file);
DIR_PARENT(dp) = parent_ndx;
DIR_FIRST_CHILD(dp) = -1;
}
if (dp)
DIR_NEXT_SIBLING(dp) = -1;
}
static void interpret_stat_error(const char *fname, int is_dir)
{
if (errno == ENOENT) {
io_error |= IOERR_VANISHED;
rprintf(FWARNING, "%s has vanished: %s\n",
is_dir ? "directory" : "file", full_fname(fname));
} else {
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, errno, "link_stat %s failed",
full_fname(fname));
}
}
/* This function is normally called by the sender, but the receiving side also
* calls it from get_dirlist() with f set to -1 so that we just construct the
* file list in memory without sending it over the wire. Also, get_dirlist()
* might call this with f set to -2, which also indicates that local filter
* rules should be ignored. */
static void send_directory(int f, struct file_list *flist, char *fbuf, int len,
int flags)
{
struct dirent *di;
unsigned remainder;
char *p;
DIR *d;
int divert_dirs = (flags & FLAG_DIVERT_DIRS) != 0;
int start = flist->used;
int filter_level = f == -2 ? SERVER_FILTERS : ALL_FILTERS;
assert(flist != NULL);
if (!(d = opendir(fbuf))) {
if (errno == ENOENT) {
if (am_sender) /* Can abuse this for vanished error w/ENOENT: */
interpret_stat_error(fbuf, True);
return;
}
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, errno, "opendir %s failed", full_fname(fbuf));
return;
}
p = fbuf + len;
if (len == 1 && *fbuf == '/')
remainder = MAXPATHLEN - 1;
else if (len < MAXPATHLEN-1) {
*p++ = '/';
*p = '\0';
remainder = MAXPATHLEN - (len + 1);
} else
remainder = 0;
for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) {
unsigned name_len;
char *dname = d_name(di);
if (dname[0] == '.' && (dname[1] == '\0'
|| (dname[1] == '.' && dname[2] == '\0')))
continue;
name_len = strlcpy(p, dname, remainder);
if (name_len >= remainder) {
char save = fbuf[len];
fbuf[len] = '\0';
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER,
"filename overflows max-path len by %u: %s/%s\n",
name_len - remainder + 1, fbuf, dname);
fbuf[len] = save;
continue;
}
if (dname[0] == '\0') {
io_error |= IOERR_GENERAL;
rprintf(FERROR_XFER,
"cannot send file with empty name in %s\n",
full_fname(fbuf));
continue;
}
send_file_name(f, flist, fbuf, NULL, flags, filter_level);
}
fbuf[len] = '\0';
if (errno) {
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, errno, "readdir(%s)", full_fname(fbuf));
}
closedir(d);
if (f >= 0 && recurse && !divert_dirs) {
int i, end = flist->used - 1;
/* send_if_directory() bumps flist->used, so use "end". */
for (i = start; i <= end; i++)
send_if_directory(f, flist, flist->files[i], fbuf, len, flags);
}
}
static void send_implied_dirs(int f, struct file_list *flist, char *fname,
char *start, char *limit, int flags, char name_type)
{
static char lastpath[MAXPATHLEN] = "";
static int lastpath_len = 0;
static struct file_struct *lastpath_struct = NULL;
struct file_struct *file;
item_list *relname_list;
relnamecache **rnpp;
int len, need_new_dir, depth = 0;
filter_rule_list save_filter_list = filter_list;
flags = (flags | FLAG_IMPLIED_DIR) & ~(FLAG_TOP_DIR | FLAG_CONTENT_DIR);
filter_list.head = filter_list.tail = NULL; /* Don't filter implied dirs. */
if (inc_recurse) {
if (lastpath_struct && F_PATHNAME(lastpath_struct) == pathname
&& lastpath_len == limit - fname
&& strncmp(lastpath, fname, lastpath_len) == 0)
need_new_dir = 0;
else
need_new_dir = 1;
} else {
char *tp = fname, *lp = lastpath;
/* Skip any initial directories in our path that we
* have in common with lastpath. */
assert(start == fname);
for ( ; ; tp++, lp++) {
if (tp == limit) {
if (*lp == '/' || *lp == '\0')
goto done;
break;
}
if (*lp != *tp)
break;
if (*tp == '/') {
start = tp;
depth++;
}
}
need_new_dir = 1;
}
if (need_new_dir) {
int save_copy_links = copy_links;
int save_xfer_dirs = xfer_dirs;
char *slash;
copy_links = xfer_dirs = 1;
*limit = '\0';
for (slash = start; (slash = strchr(slash+1, '/')) != NULL; ) {
*slash = '\0';
file = send_file_name(f, flist, fname, NULL, flags, ALL_FILTERS);
depth++;
if (!inc_recurse && file && S_ISDIR(file->mode))
change_local_filter_dir(fname, strlen(fname), depth);
*slash = '/';
}
file = send_file_name(f, flist, fname, NULL, flags, ALL_FILTERS);
if (inc_recurse) {
if (file && !S_ISDIR(file->mode))
file = NULL;
lastpath_struct = file;
} else if (file && S_ISDIR(file->mode))
change_local_filter_dir(fname, strlen(fname), ++depth);
strlcpy(lastpath, fname, sizeof lastpath);
lastpath_len = limit - fname;
*limit = '/';
copy_links = save_copy_links;
xfer_dirs = save_xfer_dirs;
if (!inc_recurse)
goto done;
}
if (!lastpath_struct)
goto done; /* dir must have vanished */
len = strlen(limit+1);
memcpy(&relname_list, F_DIR_RELNAMES_P(lastpath_struct), sizeof relname_list);
if (!relname_list) {
if (!(relname_list = new0(item_list)))
out_of_memory("send_implied_dirs");
memcpy(F_DIR_RELNAMES_P(lastpath_struct), &relname_list, sizeof relname_list);
}
rnpp = EXPAND_ITEM_LIST(relname_list, relnamecache *, 32);
if (!(*rnpp = (relnamecache*)new_array(char, sizeof (relnamecache) + len)))
out_of_memory("send_implied_dirs");
(*rnpp)->name_type = name_type;
strlcpy((*rnpp)->fname, limit+1, len + 1);
done:
filter_list = save_filter_list;
}
static NORETURN void fatal_unsafe_io_error(void)
{
/* This (sadly) can only happen when pushing data because
* the sender does not know about what kind of delete
* is in effect on the receiving side when pulling. */
rprintf(FERROR_XFER, "FATAL I/O ERROR: dying to avoid a --delete-%s issue with a pre-3.0.7 receiver.\n",
delete_during == 2 ? "delay" : "during");
exit_cleanup(RERR_UNSUPPORTED);
}
static void send1extra(int f, struct file_struct *file, struct file_list *flist)
{
char fbuf[MAXPATHLEN];
item_list *relname_list;
int len, dlen, flags = FLAG_DIVERT_DIRS | FLAG_CONTENT_DIR;
size_t j;
f_name(file, fbuf);
dlen = strlen(fbuf);
if (!change_pathname(file, NULL, 0))
exit_cleanup(RERR_FILESELECT);
change_local_filter_dir(fbuf, dlen, send_dir_depth);
if (file->flags & FLAG_CONTENT_DIR) {
if (one_file_system) {
STRUCT_STAT st;
if (link_stat(fbuf, &st, copy_dirlinks) != 0) {
interpret_stat_error(fbuf, True);
return;
}
filesystem_dev = st.st_dev;
}
send_directory(f, flist, fbuf, dlen, flags);
}
if (!relative_paths)
return;
memcpy(&relname_list, F_DIR_RELNAMES_P(file), sizeof relname_list);
if (!relname_list)
return;
for (j = 0; j < relname_list->count; j++) {
char *slash;
relnamecache *rnp = ((relnamecache**)relname_list->items)[j];
char name_type = rnp->name_type;
fbuf[dlen] = '/';
len = strlcpy(fbuf + dlen + 1, rnp->fname, sizeof fbuf - dlen - 1);
free(rnp);
if (len >= (int)sizeof fbuf)
continue; /* Impossible... */
slash = strchr(fbuf+dlen+1, '/');
if (slash) {
send_implied_dirs(f, flist, fbuf, fbuf+dlen+1, slash, flags, name_type);
continue;
}
if (name_type != NORMAL_NAME) {
STRUCT_STAT st;
if (name_type == MISSING_NAME)
memset(&st, 0, sizeof st);
else if (link_stat(fbuf, &st, 1) != 0) {
interpret_stat_error(fbuf, True);
continue;
}
send_file_name(f, flist, fbuf, &st, FLAG_TOP_DIR | flags, ALL_FILTERS);
} else
send_file_name(f, flist, fbuf, NULL, FLAG_TOP_DIR | flags, ALL_FILTERS);
}
free(relname_list);
}
void send_extra_file_list(int f, int at_least)
{
struct file_list *flist;
int64 start_write;
uint16 prev_flags;
int save_io_error = io_error;
if (flist_eof)
return;
if (at_least < 0)
at_least = file_total - file_old_total + 1;
/* Keep sending data until we have the requested number of
* files in the upcoming file-lists. */
while (file_total - file_old_total < at_least) {
struct file_struct *file = dir_flist->sorted[send_dir_ndx];
int dir_ndx, dstart = stats.num_dirs;
const char *pathname = F_PATHNAME(file);
int32 *dp;
flist = flist_new(0, "send_extra_file_list");
start_write = stats.total_written;
if (unsort_ndx)
dir_ndx = F_NDX(file);
else
dir_ndx = send_dir_ndx;
write_ndx(f, NDX_FLIST_OFFSET - dir_ndx);
flist->parent_ndx = dir_ndx;
send1extra(f, file, flist);
prev_flags = file->flags;
dp = F_DIR_NODE_P(file);
/* If there are any duplicate directory names that follow, we
* send all the dirs together in one file-list. The dir_flist
* tree links all the child subdirs onto the last dup dir. */
while ((dir_ndx = DIR_NEXT_SIBLING(dp)) >= 0
&& dir_flist->sorted[dir_ndx]->flags & FLAG_DUPLICATE) {
send_dir_ndx = dir_ndx;
file = dir_flist->sorted[dir_ndx];
/* Try to avoid some duplicate scanning of identical dirs. */
if (F_PATHNAME(file) == pathname && prev_flags & FLAG_CONTENT_DIR)
file->flags &= ~FLAG_CONTENT_DIR;
send1extra(f, file, flist);
prev_flags = file->flags;
dp = F_DIR_NODE_P(file);
}
if (io_error == save_io_error || ignore_errors)
write_byte(f, 0);
else if (use_safe_inc_flist) {
write_shortint(f, XMIT_EXTENDED_FLAGS|XMIT_IO_ERROR_ENDLIST);
write_varint(f, io_error);
} else {
if (delete_during)
fatal_unsafe_io_error();
write_byte(f, 0);
}
if (need_unsorted_flist) {
if (!(flist->sorted = new_array(struct file_struct *, flist->used)))
out_of_memory("send_extra_file_list");
memcpy(flist->sorted, flist->files,
flist->used * sizeof (struct file_struct*));
} else
flist->sorted = flist->files;
flist_sort_and_clean(flist, 0);
add_dirs_to_tree(send_dir_ndx, flist, stats.num_dirs - dstart);
flist_done_allocating(flist);
file_total += flist->used;
stats.flist_size += stats.total_written - start_write;
stats.num_files += flist->used;
if (DEBUG_GTE(FLIST, 3))
output_flist(flist);
if (DIR_FIRST_CHILD(dp) >= 0) {
send_dir_ndx = DIR_FIRST_CHILD(dp);
send_dir_depth++;
} else {
while (DIR_NEXT_SIBLING(dp) < 0) {
if ((send_dir_ndx = DIR_PARENT(dp)) < 0) {
write_ndx(f, NDX_FLIST_EOF);
flist_eof = 1;
if (DEBUG_GTE(FLIST, 3))
rprintf(FINFO, "[%s] flist_eof=1\n", who_am_i());
change_local_filter_dir(NULL, 0, 0);
goto finish;
}
send_dir_depth--;
file = dir_flist->sorted[send_dir_ndx];
dp = F_DIR_NODE_P(file);
}
send_dir_ndx = DIR_NEXT_SIBLING(dp);
}
}
finish:
if (io_error != save_io_error && protocol_version == 30 && !ignore_errors)
send_msg_int(MSG_IO_ERROR, io_error);
}
struct file_list *send_file_list(int f, int argc, char *argv[])
{
static const char *lastdir;
static int lastdir_len = -1;
int len, dirlen;
STRUCT_STAT st;
char *p, *dir;
struct file_list *flist;
struct timeval start_tv, end_tv;
int64 start_write;
int use_ff_fd = 0;
int disable_buffering, reenable_multiplex = -1;
int flags = recurse ? FLAG_CONTENT_DIR : 0;
int reading_remotely = filesfrom_host != NULL;
int rl_flags = (reading_remotely ? 0 : RL_DUMP_COMMENTS)
#ifdef ICONV_OPTION
| (filesfrom_convert ? RL_CONVERT : 0)
#endif
| (eol_nulls || reading_remotely ? RL_EOL_NULLS : 0);
int implied_dot_dir = 0;
rprintf(FLOG, "building file list\n");
if (show_filelist_p())
start_filelist_progress("building file list");
else if (inc_recurse && INFO_GTE(FLIST, 1) && !am_server)
rprintf(FCLIENT, "sending incremental file list\n");
start_write = stats.total_written;
gettimeofday(&start_tv, NULL);
if (relative_paths && protocol_version >= 30)
implied_dirs = 1; /* We send flagged implied dirs */
#ifdef SUPPORT_HARD_LINKS
if (preserve_hard_links && protocol_version >= 30 && !cur_flist)
init_hard_links();
#endif
flist = cur_flist = flist_new(0, "send_file_list");
if (inc_recurse) {
dir_flist = flist_new(FLIST_TEMP, "send_file_list");
flags |= FLAG_DIVERT_DIRS;
} else
dir_flist = cur_flist;
disable_buffering = io_start_buffering_out(f);
if (filesfrom_fd >= 0) {
if (argv[0] && !change_dir(argv[0], CD_NORMAL)) {
rsyserr(FERROR_XFER, errno, "change_dir %s failed",
full_fname(argv[0]));
exit_cleanup(RERR_FILESELECT);
}
if (protocol_version < 31) {
/* Older protocols send the files-from data w/o packaging
* it in multiplexed I/O packets, so temporarily switch
* to buffered I/O to match this behavior. */
reenable_multiplex = io_end_multiplex_in(MPLX_TO_BUFFERED);
}
use_ff_fd = 1;
}
if (!orig_dir)
orig_dir = strdup(curr_dir);
while (1) {
char fbuf[MAXPATHLEN], *fn, name_type;
if (use_ff_fd) {
if (read_line(filesfrom_fd, fbuf, sizeof fbuf, rl_flags) == 0)
break;
sanitize_path(fbuf, fbuf, "", 0, SP_KEEP_DOT_DIRS);
} else {
if (argc-- == 0)
break;
strlcpy(fbuf, *argv++, MAXPATHLEN);
if (sanitize_paths)
sanitize_path(fbuf, fbuf, "", 0, SP_KEEP_DOT_DIRS);
}
len = strlen(fbuf);
if (relative_paths) {
/* We clean up fbuf below. */
name_type = NORMAL_NAME;
} else if (!len || fbuf[len - 1] == '/') {
if (len == 2 && fbuf[0] == '.') {
/* Turn "./" into just "." rather than "./." */
fbuf[--len] = '\0';
} else {
if (len + 1 >= MAXPATHLEN)
overflow_exit("send_file_list");
fbuf[len++] = '.';
fbuf[len] = '\0';
}
name_type = DOTDIR_NAME;
} else if (len > 1 && fbuf[len-1] == '.' && fbuf[len-2] == '.'
&& (len == 2 || fbuf[len-3] == '/')) {
if (len + 2 >= MAXPATHLEN)
overflow_exit("send_file_list");
fbuf[len++] = '/';
fbuf[len++] = '.';
fbuf[len] = '\0';
name_type = DOTDIR_NAME;
} else if (fbuf[len-1] == '.' && (len == 1 || fbuf[len-2] == '/'))
name_type = DOTDIR_NAME;
else
name_type = NORMAL_NAME;
dir = NULL;
if (!relative_paths) {
p = strrchr(fbuf, '/');
if (p) {
*p = '\0';
if (p == fbuf)
dir = "/";
else
dir = fbuf;
len -= p - fbuf + 1;
fn = p + 1;
} else
fn = fbuf;
} else {
if ((p = strstr(fbuf, "/./")) != NULL) {
*p = '\0';
if (p == fbuf)
dir = "/";
else {
dir = fbuf;
clean_fname(dir, 0);
}
fn = p + 3;
while (*fn == '/')
fn++;
if (!*fn)
*--fn = '\0'; /* ensure room for '.' */
} else
fn = fbuf;
/* A leading ./ can be used in relative mode to affect
* the dest dir without its name being in the path. */
if (*fn == '.' && fn[1] == '/' && fn[2] && !implied_dot_dir)
implied_dot_dir = -1;
len = clean_fname(fn, CFN_KEEP_TRAILING_SLASH
| CFN_DROP_TRAILING_DOT_DIR);
if (len == 1) {
if (fn[0] == '/') {
fn = "/.";
len = 2;
name_type = DOTDIR_NAME;
} else if (fn[0] == '.')
name_type = DOTDIR_NAME;
} else if (fn[len-1] == '/') {
fn[--len] = '\0';
if (len == 1 && *fn == '.')
name_type = DOTDIR_NAME;
else
name_type = SLASH_ENDING_NAME;
}
/* Reject a ".." dir in the active part of the path. */
for (p = fn; (p = strstr(p, "..")) != NULL; p += 2) {
if ((p[2] == '/' || p[2] == '\0')
&& (p == fn || p[-1] == '/')) {
rprintf(FERROR,
"found \"..\" dir in relative path: %s\n",
fn);
exit_cleanup(RERR_SYNTAX);
}
}
}
if (!*fn) {
len = 1;
fn = ".";
name_type = DOTDIR_NAME;
}
dirlen = dir ? strlen(dir) : 0;
if (dirlen != lastdir_len || memcmp(lastdir, dir, dirlen) != 0) {
if (!change_pathname(NULL, dir, -dirlen))
goto bad_path;
lastdir = pathname;
lastdir_len = pathname_len;
} else if (!change_pathname(NULL, lastdir, lastdir_len)) {
bad_path:
if (implied_dot_dir < 0)
implied_dot_dir = 0;
continue;
}
if (implied_dot_dir < 0) {
implied_dot_dir = 1;
send_file_name(f, flist, ".", NULL, (flags | FLAG_IMPLIED_DIR) & ~FLAG_CONTENT_DIR, ALL_FILTERS);
}
if (fn != fbuf)
memmove(fbuf, fn, len + 1);
if (link_stat(fbuf, &st, copy_dirlinks || name_type != NORMAL_NAME) != 0
|| (name_type != DOTDIR_NAME && is_daemon_excluded(fbuf, S_ISDIR(st.st_mode)))
|| (relative_paths && path_is_daemon_excluded(fbuf, 1))) {
if (errno != ENOENT || missing_args == 0) {
/* This is a transfer error, but inhibit deletion
* only if we might be omitting an existing file. */
if (errno != ENOENT)
io_error |= IOERR_GENERAL;
rsyserr(FERROR_XFER, errno, "link_stat %s failed",
full_fname(fbuf));
continue;
} else if (missing_args == 1) {
/* Just ignore the arg. */
continue;
} else /* (missing_args == 2) */ {
/* Send the arg as a "missing" entry with
* mode 0, which tells the generator to delete it. */
memset(&st, 0, sizeof st);
}
}
/* A dot-dir should not be excluded! */
if (name_type != DOTDIR_NAME && st.st_mode != 0
&& is_excluded(fbuf, S_ISDIR(st.st_mode) != 0, ALL_FILTERS))
continue;
if (S_ISDIR(st.st_mode) && !xfer_dirs) {
rprintf(FINFO, "skipping directory %s\n", fbuf);
continue;
}
if (inc_recurse && relative_paths && *fbuf) {
if ((p = strchr(fbuf+1, '/')) != NULL) {
if (p - fbuf == 1 && *fbuf == '.') {
if ((fn = strchr(p+1, '/')) != NULL)
p = fn;
} else
fn = p;
send_implied_dirs(f, flist, fbuf, fbuf, p, flags,
IS_MISSING_FILE(st) ? MISSING_NAME : name_type);
if (fn == p)
continue;
}
} else if (implied_dirs && (p=strrchr(fbuf,'/')) && p != fbuf) {
/* Send the implied directories at the start of the
* source spec, so we get their permissions right. */
send_implied_dirs(f, flist, fbuf, fbuf, p, flags, 0);
}
if (one_file_system)
filesystem_dev = st.st_dev;
if (recurse || (xfer_dirs && name_type != NORMAL_NAME)) {
struct file_struct *file;
file = send_file_name(f, flist, fbuf, &st,
FLAG_TOP_DIR | FLAG_CONTENT_DIR | flags,
NO_FILTERS);
if (!file)
continue;
if (inc_recurse) {
if (name_type == DOTDIR_NAME) {
if (send_dir_depth < 0) {
send_dir_depth = 0;
change_local_filter_dir(fbuf, len, send_dir_depth);
}
send_directory(f, flist, fbuf, len, flags);
}
} else
send_if_directory(f, flist, file, fbuf, len, flags);
} else
send_file_name(f, flist, fbuf, &st, flags, NO_FILTERS);
}
if (reenable_multiplex >= 0)
io_start_multiplex_in(reenable_multiplex);
gettimeofday(&end_tv, NULL);
stats.flist_buildtime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
+ (end_tv.tv_usec - start_tv.tv_usec) / 1000;
if (stats.flist_buildtime == 0)
stats.flist_buildtime = 1;
start_tv = end_tv;
/* Indicate end of file list */
if (io_error == 0 || ignore_errors)
write_byte(f, 0);
else if (use_safe_inc_flist) {
write_shortint(f, XMIT_EXTENDED_FLAGS|XMIT_IO_ERROR_ENDLIST);
write_varint(f, io_error);
} else {
if (delete_during && inc_recurse)
fatal_unsafe_io_error();
write_byte(f, 0);
}
#ifdef SUPPORT_HARD_LINKS
if (preserve_hard_links && protocol_version >= 30 && !inc_recurse)
idev_destroy();
#endif
if (show_filelist_p())
finish_filelist_progress(flist);
gettimeofday(&end_tv, NULL);
stats.flist_xfertime = (int64)(end_tv.tv_sec - start_tv.tv_sec) * 1000
+ (end_tv.tv_usec - start_tv.tv_usec) / 1000;
/* When converting names, both sides keep an unsorted file-list array
* because the names will differ on the sending and receiving sides
* (both sides will use the unsorted index number for each item). */
/* Sort the list without removing any duplicates. This allows the
* receiving side to ask for whatever name it kept. For incremental
* recursion mode, the sender marks duplicate dirs so that it can
* send them together in a single file-list. */
if (need_unsorted_flist) {
if (!(flist->sorted = new_array(struct file_struct *, flist->used)))
out_of_memory("send_file_list");
memcpy(flist->sorted, flist->files,
flist->used * sizeof (struct file_struct*));
} else
flist->sorted = flist->files;
flist_sort_and_clean(flist, 0);
file_total += flist->used;
file_old_total += flist->used;
if (numeric_ids <= 0 && !inc_recurse)
send_id_list(f);
/* send the io_error flag */
if (protocol_version < 30)
write_int(f, ignore_errors ? 0 : io_error);
else if (!use_safe_inc_flist && io_error && !ignore_errors)
send_msg_int(MSG_IO_ERROR, io_error);
if (disable_buffering)
io_end_buffering_out(IOBUF_FREE_BUFS);
stats.flist_size = stats.total_written - start_write;
stats.num_files = flist->used;
if (DEBUG_GTE(FLIST, 3))
output_flist(flist);
if (DEBUG_GTE(FLIST, 2))
rprintf(FINFO, "send_file_list done\n");
if (inc_recurse) {
send_dir_depth = 1;
add_dirs_to_tree(-1, flist, stats.num_dirs);
if (!file_total || strcmp(flist->sorted[flist->low]->basename, ".") != 0)
flist->parent_ndx = -1;
flist_done_allocating(flist);
if (send_dir_ndx < 0) {
write_ndx(f, NDX_FLIST_EOF);
flist_eof = 1;
if (DEBUG_GTE(FLIST, 3))
rprintf(FINFO, "[%s] flist_eof=1\n", who_am_i());
}
else if (file_total == 1) {
/* If we're creating incremental file-lists and there
* was just 1 item in the first file-list, send 1 more
* file-list to check if this is a 1-file xfer. */
send_extra_file_list(f, 1);
}
} else {
flist_eof = 1;
if (DEBUG_GTE(FLIST, 3))
rprintf(FINFO, "[%s] flist_eof=1\n", who_am_i());
}
return flist;
}
struct file_list *recv_file_list(int f)
{
struct file_list *flist;
int dstart, flags;
int64 start_read;
if (!first_flist) {
if (show_filelist_p())
start_filelist_progress("receiving file list");
else if (inc_recurse && INFO_GTE(FLIST, 1) && !am_server)
rprintf(FCLIENT, "receiving incremental file list\n");
rprintf(FLOG, "receiving file list\n");
if (usermap)
parse_name_map(usermap, True);
if (groupmap)
parse_name_map(groupmap, False);
}
start_read = stats.total_read;
#ifdef SUPPORT_HARD_LINKS
if (preserve_hard_links && !first_flist)
init_hard_links();
#endif
flist = flist_new(0, "recv_file_list");
if (inc_recurse) {
if (flist->ndx_start == 1)
dir_flist = flist_new(FLIST_TEMP, "recv_file_list");
dstart = dir_flist->used;
} else {
dir_flist = flist;
dstart = 0;
}
while ((flags = read_byte(f)) != 0) {
struct file_struct *file;
if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS))
flags |= read_byte(f) << 8;
if (flags == (XMIT_EXTENDED_FLAGS|XMIT_IO_ERROR_ENDLIST)) {
int err;
if (!use_safe_inc_flist) {
rprintf(FERROR, "Invalid flist flag: %x\n", flags);
exit_cleanup(RERR_PROTOCOL);
}
err = read_varint(f);
if (!ignore_errors)
io_error |= err;
break;
}
flist_expand(flist, 1);
file = recv_file_entry(f, flist, flags);
if (S_ISREG(file->mode)) {
/* Already counted */
} else if (S_ISDIR(file->mode)) {
if (inc_recurse) {
flist_expand(dir_flist, 1);
dir_flist->files[dir_flist->used++] = file;
}
stats.num_dirs++;
} else if (S_ISLNK(file->mode))
stats.num_symlinks++;
else if (IS_DEVICE(file->mode))
stats.num_symlinks++;
else
stats.num_specials++;
flist->files[flist->used++] = file;
maybe_emit_filelist_progress(flist->used);
if (DEBUG_GTE(FLIST, 2)) {
char *name = f_name(file, NULL);
rprintf(FINFO, "recv_file_name(%s)\n", NS(name));
}
}
file_total += flist->used;
if (DEBUG_GTE(FLIST, 2))
rprintf(FINFO, "received %d names\n", flist->used);
if (show_filelist_p())
finish_filelist_progress(flist);
if (need_unsorted_flist) {
/* Create an extra array of index pointers that we can sort for
* the generator's use (for wading through the files in sorted
* order and for calling flist_find()). We keep the "files"
* list unsorted for our exchange of index numbers with the
* other side (since their names may not sort the same). */
if (!(flist->sorted = new_array(struct file_struct *, flist->used)))
out_of_memory("recv_file_list");
memcpy(flist->sorted, flist->files,
flist->used * sizeof (struct file_struct*));
if (inc_recurse && dir_flist->used > dstart) {
static int dir_flist_malloced = 0;
if (dir_flist_malloced < dir_flist->malloced) {
dir_flist->sorted = realloc_array(dir_flist->sorted,
struct file_struct *,
dir_flist->malloced);
dir_flist_malloced = dir_flist->malloced;
}
memcpy(dir_flist->sorted + dstart, dir_flist->files + dstart,
(dir_flist->used - dstart) * sizeof (struct file_struct*));
fsort(dir_flist->sorted + dstart, dir_flist->used - dstart);
}
} else {
flist->sorted = flist->files;
if (inc_recurse && dir_flist->used > dstart) {
dir_flist->sorted = dir_flist->files;
fsort(dir_flist->sorted + dstart, dir_flist->used - dstart);
}
}
if (inc_recurse)
flist_done_allocating(flist);
else if (f >= 0) {
recv_id_list(f, flist);
flist_eof = 1;
if (DEBUG_GTE(FLIST, 3))
rprintf(FINFO, "[%s] flist_eof=1\n", who_am_i());
}
/* The --relative option sends paths with a leading slash, so we need
* to specify the strip_root option here. We rejected leading slashes
* for a non-relative transfer in recv_file_entry(). */
flist_sort_and_clean(flist, relative_paths);
if (protocol_version < 30) {
/* Recv the io_error flag */
int err = read_int(f);
if (!ignore_errors)
io_error |= err;
} else if (inc_recurse && flist->ndx_start == 1) {
if (!file_total || strcmp(flist->sorted[flist->low]->basename, ".") != 0)
flist->parent_ndx = -1;
}
if (DEBUG_GTE(FLIST, 3))
output_flist(flist);
if (DEBUG_GTE(FLIST, 2))
rprintf(FINFO, "recv_file_list done\n");
stats.flist_size += stats.total_read - start_read;
stats.num_files += flist->used;
return flist;
}
/* This is only used once by the receiver if the very first file-list
* has exactly one item in it. */
void recv_additional_file_list(int f)
{
struct file_list *flist;
int ndx = read_ndx(f);
if (ndx == NDX_FLIST_EOF) {
flist_eof = 1;
if (DEBUG_GTE(FLIST, 3))
rprintf(FINFO, "[%s] flist_eof=1\n", who_am_i());
change_local_filter_dir(NULL, 0, 0);
} else {
ndx = NDX_FLIST_OFFSET - ndx;
if (ndx < 0 || ndx >= dir_flist->used) {
ndx = NDX_FLIST_OFFSET - ndx;
rprintf(FERROR,
"[%s] Invalid dir index: %d (%d - %d)\n",
who_am_i(), ndx, NDX_FLIST_OFFSET,
NDX_FLIST_OFFSET - dir_flist->used + 1);
exit_cleanup(RERR_PROTOCOL);
}
if (DEBUG_GTE(FLIST, 3)) {
rprintf(FINFO, "[%s] receiving flist for dir %d\n",
who_am_i(), ndx);
}
flist = recv_file_list(f);
flist->parent_ndx = ndx;
}
}
/* Search for an identically-named item in the file list. Note that the
* items must agree in their directory-ness, or no match is returned. */
int flist_find(struct file_list *flist, struct file_struct *f)
{
int low = flist->low, high = flist->high;
int diff, mid, mid_up;
while (low <= high) {
mid = (low + high) / 2;
if (F_IS_ACTIVE(flist->sorted[mid]))
mid_up = mid;
else {
/* Scan for the next non-empty entry using the cached
* distance values. If the value isn't fully up-to-
* date, update it. */
mid_up = mid + F_DEPTH(flist->sorted[mid]);
if (!F_IS_ACTIVE(flist->sorted[mid_up])) {
do {
mid_up += F_DEPTH(flist->sorted[mid_up]);
} while (!F_IS_ACTIVE(flist->sorted[mid_up]));
F_DEPTH(flist->sorted[mid]) = mid_up - mid;
}
if (mid_up > high) {
/* If there's nothing left above us, set high to
* a non-empty entry below us and continue. */
high = mid - (int)flist->sorted[mid]->len32;
if (!F_IS_ACTIVE(flist->sorted[high])) {
do {
high -= (int)flist->sorted[high]->len32;
} while (!F_IS_ACTIVE(flist->sorted[high]));
flist->sorted[mid]->len32 = mid - high;
}
continue;
}
}
diff = f_name_cmp(flist->sorted[mid_up], f);
if (diff == 0) {
if (protocol_version < 29
&& S_ISDIR(flist->sorted[mid_up]->mode)
!= S_ISDIR(f->mode))
return -1;
return mid_up;
}
if (diff < 0)
low = mid_up + 1;
else
high = mid - 1;
}
return -1;
}
/* Search for an identically-named item in the file list. Differs from
* flist_find in that an item that agrees with "f" in directory-ness is
* preferred but one that does not is still found. */
int flist_find_ignore_dirness(struct file_list *flist, struct file_struct *f)
{
mode_t save_mode;
int ndx;
/* First look for an item that agrees in directory-ness. */
ndx = flist_find(flist, f);
if (ndx >= 0)
return ndx;
/* Temporarily flip f->mode to look for an item of opposite
* directory-ness. */
save_mode = f->mode;
f->mode = S_ISDIR(f->mode) ? S_IFREG : S_IFDIR;
ndx = flist_find(flist, f);
f->mode = save_mode;
return ndx;
}
/*
* Free up any resources a file_struct has allocated
* and clear the file.
*/
void clear_file(struct file_struct *file)
{
/* The +1 zeros out the first char of the basename. */
memset(file, 0, FILE_STRUCT_LEN + 1);
/* In an empty entry, F_DEPTH() is an offset to the next non-empty
* entry. Likewise for len32 in the opposite direction. We assume
* that we're alone for now since flist_find() will adjust the counts
* it runs into that aren't up-to-date. */
file->len32 = F_DEPTH(file) = 1;
}
/* Allocate a new file list. */
struct file_list *flist_new(int flags, char *msg)
{
struct file_list *flist;
if (!(flist = new0(struct file_list)))
out_of_memory(msg);
if (flags & FLIST_TEMP) {
if (!(flist->file_pool = pool_create(SMALL_EXTENT, 0,
out_of_memory,
POOL_INTERN)))
out_of_memory(msg);
} else {
/* This is a doubly linked list with prev looping back to
* the end of the list, but the last next pointer is NULL. */
if (!first_flist) {
flist->file_pool = pool_create(NORMAL_EXTENT, 0,
out_of_memory,
POOL_INTERN);
if (!flist->file_pool)
out_of_memory(msg);
flist->ndx_start = flist->flist_num = inc_recurse ? 1 : 0;
first_flist = cur_flist = flist->prev = flist;
} else {
struct file_list *prev = first_flist->prev;
flist->file_pool = first_flist->file_pool;
flist->ndx_start = prev->ndx_start + prev->used + 1;
flist->flist_num = prev->flist_num + 1;
flist->prev = prev;
prev->next = first_flist->prev = flist;
}
flist->pool_boundary = pool_boundary(flist->file_pool, 0);
flist_cnt++;
}
return flist;
}
/* Free up all elements in a flist. */
void flist_free(struct file_list *flist)
{
if (!flist->prev) {
/* Was FLIST_TEMP dir-list. */
} else if (flist == flist->prev) {
first_flist = cur_flist = NULL;
file_total = 0;
flist_cnt = 0;
} else {
if (flist == cur_flist)
cur_flist = flist->next;
if (flist == first_flist)
first_flist = first_flist->next;
else {
flist->prev->next = flist->next;
if (!flist->next)
flist->next = first_flist;
}
flist->next->prev = flist->prev;
file_total -= flist->used;
flist_cnt--;
}
if (!flist->prev || !flist_cnt)
pool_destroy(flist->file_pool);
else
pool_free_old(flist->file_pool, flist->pool_boundary);
if (flist->sorted && flist->sorted != flist->files)
free(flist->sorted);
free(flist->files);
free(flist);
}
/* This routine ensures we don't have any duplicate names in our file list.
* duplicate names can cause corruption because of the pipelining. */
static void flist_sort_and_clean(struct file_list *flist, int strip_root)
{
char fbuf[MAXPATHLEN];
int i, prev_i;
if (!flist)
return;
if (flist->used == 0) {
flist->high = -1;
flist->low = 0;
return;
}
fsort(flist->sorted, flist->used);
if (!am_sender || inc_recurse) {
for (i = prev_i = 0; i < flist->used; i++) {
if (F_IS_ACTIVE(flist->sorted[i])) {
prev_i = i;
break;
}
}
flist->low = prev_i;
} else {
i = prev_i = flist->used - 1;
flist->low = 0;
}
while (++i < flist->used) {
int j;
struct file_struct *file = flist->sorted[i];
if (!F_IS_ACTIVE(file))
continue;
if (f_name_cmp(file, flist->sorted[prev_i]) == 0)
j = prev_i;
else if (protocol_version >= 29 && S_ISDIR(file->mode)) {
int save_mode = file->mode;
/* Make sure that this directory doesn't duplicate a
* non-directory earlier in the list. */
flist->high = prev_i;
file->mode = S_IFREG;
j = flist_find(flist, file);
file->mode = save_mode;
} else
j = -1;
if (j >= 0) {
int keep, drop;
/* If one is a dir and the other is not, we want to
* keep the dir because it might have contents in the
* list. Otherwise keep the first one. */
if (S_ISDIR(file->mode)) {
struct file_struct *fp = flist->sorted[j];
if (!S_ISDIR(fp->mode))
keep = i, drop = j;
else {
if (am_sender)
file->flags |= FLAG_DUPLICATE;
else { /* Make sure we merge our vital flags. */
fp->flags |= file->flags & (FLAG_TOP_DIR|FLAG_CONTENT_DIR);
fp->flags &= file->flags | ~FLAG_IMPLIED_DIR;
}
keep = j, drop = i;
}
} else
keep = j, drop = i;
if (!am_sender) {
if (DEBUG_GTE(DUP, 1)) {
rprintf(FINFO,
"removing duplicate name %s from file list (%d)\n",
f_name(file, fbuf), drop + flist->ndx_start);
}
clear_file(flist->sorted[drop]);
}
if (keep == i) {
if (flist->low == drop) {
for (j = drop + 1;
j < i && !F_IS_ACTIVE(flist->sorted[j]);
j++) {}
flist->low = j;
}
prev_i = i;
}
} else
prev_i = i;
}
flist->high = prev_i;
if (strip_root) {
/* We need to strip off the leading slashes for relative
* paths, but this must be done _after_ the sorting phase. */
for (i = flist->low; i <= flist->high; i++) {
struct file_struct *file = flist->sorted[i];
if (!file->dirname)
continue;
while (*file->dirname == '/')
file->dirname++;
if (!*file->dirname)
file->dirname = NULL;
}
}
if (prune_empty_dirs && !am_sender) {
int j, prev_depth = 0;
prev_i = 0; /* It's OK that this isn't really true. */
for (i = flist->low; i <= flist->high; i++) {
struct file_struct *fp, *file = flist->sorted[i];
/* This temporarily abuses the F_DEPTH() value for a
* directory that is in a chain that might get pruned.
* We restore the old value if it gets a reprieve. */
if (S_ISDIR(file->mode) && F_DEPTH(file)) {
/* Dump empty dirs when coming back down. */
for (j = prev_depth; j >= F_DEPTH(file); j--) {
fp = flist->sorted[prev_i];
if (F_DEPTH(fp) >= 0)
break;
prev_i = -F_DEPTH(fp)-1;
clear_file(fp);
}
prev_depth = F_DEPTH(file);
if (is_excluded(f_name(file, fbuf), 1,
ALL_FILTERS)) {
/* Keep dirs through this dir. */
for (j = prev_depth-1; ; j--) {
fp = flist->sorted[prev_i];
if (F_DEPTH(fp) >= 0)
break;
prev_i = -F_DEPTH(fp)-1;
F_DEPTH(fp) = j;
}
} else
F_DEPTH(file) = -prev_i-1;
prev_i = i;
} else {
/* Keep dirs through this non-dir. */
for (j = prev_depth; ; j--) {
fp = flist->sorted[prev_i];
if (F_DEPTH(fp) >= 0)
break;
prev_i = -F_DEPTH(fp)-1;
F_DEPTH(fp) = j;
}
}
}
/* Dump all remaining empty dirs. */
while (1) {
struct file_struct *fp = flist->sorted[prev_i];
if (F_DEPTH(fp) >= 0)
break;
prev_i = -F_DEPTH(fp)-1;
clear_file(fp);
}
for (i = flist->low; i <= flist->high; i++) {
if (F_IS_ACTIVE(flist->sorted[i]))
break;
}
flist->low = i;
for (i = flist->high; i >= flist->low; i--) {
if (F_IS_ACTIVE(flist->sorted[i]))
break;
}
flist->high = i;
}
}
static void output_flist(struct file_list *flist)
{
char uidbuf[16], gidbuf[16], depthbuf[16];
struct file_struct *file;
const char *root, *dir, *slash, *name, *trail;
const char *who = who_am_i();
int i;
rprintf(FINFO, "[%s] flist start=%d, used=%d, low=%d, high=%d\n",
who, flist->ndx_start, flist->used, flist->low, flist->high);
for (i = 0; i < flist->used; i++) {
file = flist->files[i];
if ((am_root || am_sender) && uid_ndx) {
snprintf(uidbuf, sizeof uidbuf, " uid=%u",
F_OWNER(file));
} else
*uidbuf = '\0';
if (gid_ndx) {
static char parens[] = "(\0)\0\0\0";
char *pp = parens + (file->flags & FLAG_SKIP_GROUP ? 0 : 3);
snprintf(gidbuf, sizeof gidbuf, " gid=%s%u%s",
pp, F_GROUP(file), pp + 2);
} else
*gidbuf = '\0';
if (!am_sender)
snprintf(depthbuf, sizeof depthbuf, "%d", F_DEPTH(file));
if (F_IS_ACTIVE(file)) {
root = am_sender ? NS(F_PATHNAME(file)) : depthbuf;
if ((dir = file->dirname) == NULL)
dir = slash = "";
else
slash = "/";
name = file->basename;
trail = S_ISDIR(file->mode) ? "/" : "";
} else
root = dir = slash = name = trail = "";
rprintf(FINFO,
"[%s] i=%d %s %s%s%s%s mode=0%o len=%s%s%s flags=%x\n",
who, i + flist->ndx_start,
root, dir, slash, name, trail,
(int)file->mode, comma_num(F_LENGTH(file)),
uidbuf, gidbuf, file->flags);
}
}
enum fnc_state { s_DIR, s_SLASH, s_BASE, s_TRAILING };
enum fnc_type { t_PATH, t_ITEM };
static int found_prefix;
/* Compare the names of two file_struct entities, similar to how strcmp()
* would do if it were operating on the joined strings.
*
* Some differences beginning with protocol_version 29: (1) directory names
* are compared with an assumed trailing slash so that they compare in a
* way that would cause them to sort immediately prior to any content they
* may have; (2) a directory of any name compares after a non-directory of
* any name at the same depth; (3) a directory with name "." compares prior
* to anything else. These changes mean that a directory and a non-dir
* with the same name will not compare as equal (protocol_version >= 29).
*
* The dirname component can be an empty string, but the basename component
* cannot (and never is in the current codebase). The basename component
* may be NULL (for a removed item), in which case it is considered to be
* after any existing item. */
int f_name_cmp(const struct file_struct *f1, const struct file_struct *f2)
{
int dif;
const uchar *c1, *c2;
enum fnc_state state1, state2;
enum fnc_type type1, type2;
enum fnc_type t_path = protocol_version >= 29 ? t_PATH : t_ITEM;
if (!f1 || !F_IS_ACTIVE(f1)) {
if (!f2 || !F_IS_ACTIVE(f2))
return 0;
return -1;
}
if (!f2 || !F_IS_ACTIVE(f2))
return 1;
c1 = (uchar*)f1->dirname;
c2 = (uchar*)f2->dirname;
if (c1 == c2)
c1 = c2 = NULL;
if (!c1) {
type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
c1 = (const uchar*)f1->basename;
if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
type1 = t_ITEM;
state1 = s_TRAILING;
c1 = (uchar*)"";
} else
state1 = s_BASE;
} else {
type1 = t_path;
state1 = s_DIR;
}
if (!c2) {
type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
c2 = (const uchar*)f2->basename;
if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
type2 = t_ITEM;
state2 = s_TRAILING;
c2 = (uchar*)"";
} else
state2 = s_BASE;
} else {
type2 = t_path;
state2 = s_DIR;
}
if (type1 != type2)
return type1 == t_PATH ? 1 : -1;
do {
if (!*c1) {
switch (state1) {
case s_DIR:
state1 = s_SLASH;
c1 = (uchar*)"/";
break;
case s_SLASH:
type1 = S_ISDIR(f1->mode) ? t_path : t_ITEM;
c1 = (const uchar*)f1->basename;
if (type1 == t_PATH && *c1 == '.' && !c1[1]) {
type1 = t_ITEM;
state1 = s_TRAILING;
c1 = (uchar*)"";
} else
state1 = s_BASE;
break;
case s_BASE:
state1 = s_TRAILING;
if (type1 == t_PATH) {
c1 = (uchar*)"/";
break;
}
/* FALL THROUGH */
case s_TRAILING:
type1 = t_ITEM;
break;
}
if (*c2 && type1 != type2)
return type1 == t_PATH ? 1 : -1;
}
if (!*c2) {
switch (state2) {
case s_DIR:
state2 = s_SLASH;
c2 = (uchar*)"/";
break;
case s_SLASH:
type2 = S_ISDIR(f2->mode) ? t_path : t_ITEM;
c2 = (const uchar*)f2->basename;
if (type2 == t_PATH && *c2 == '.' && !c2[1]) {
type2 = t_ITEM;
state2 = s_TRAILING;
c2 = (uchar*)"";
} else
state2 = s_BASE;
break;
case s_BASE:
state2 = s_TRAILING;
if (type2 == t_PATH) {
c2 = (uchar*)"/";
break;
}
/* FALL THROUGH */
case s_TRAILING:
found_prefix = 1;
if (!*c1)
return 0;
type2 = t_ITEM;
break;
}
if (type1 != type2)
return type1 == t_PATH ? 1 : -1;
}
} while ((dif = (int)*c1++ - (int)*c2++) == 0);
return dif;
}
/* Returns 1 if f1's filename has all of f2's filename as a prefix. This does
* not match if f2's basename is not an exact match of a path element in f1.
* E.g. /path/foo is not a prefix of /path/foobar/baz, but /path/foobar is. */
int f_name_has_prefix(const struct file_struct *f1, const struct file_struct *f2)
{
found_prefix = 0;
f_name_cmp(f1, f2);
return found_prefix;
}
char *f_name_buf(void)
{
static char names[5][MAXPATHLEN];
static unsigned int n;
n = (n + 1) % (sizeof names / sizeof names[0]);
return names[n];
}
/* Return a copy of the full filename of a flist entry, using the indicated
* buffer or one of 5 static buffers if fbuf is NULL. No size-checking is
* done because we checked the size when creating the file_struct entry.
*/
char *f_name(const struct file_struct *f, char *fbuf)
{
if (!f || !F_IS_ACTIVE(f))
return NULL;
if (!fbuf)
fbuf = f_name_buf();
if (f->dirname) {
int len = strlen(f->dirname);
memcpy(fbuf, f->dirname, len);
fbuf[len] = '/';
strlcpy(fbuf + len + 1, f->basename, MAXPATHLEN - (len + 1));
} else
strlcpy(fbuf, f->basename, MAXPATHLEN);
return fbuf;
}
/* Do a non-recursive scan of the named directory, possibly ignoring all
* exclude rules except for the daemon's. If "dlen" is >=0, it is the length
* of the dirname string, and also indicates that "dirname" is a MAXPATHLEN
* buffer (the functions we call will append names onto the end, but the old
* dir value will be restored on exit). */
struct file_list *get_dirlist(char *dirname, int dlen, int flags)
{
struct file_list *dirlist;
char dirbuf[MAXPATHLEN];
int save_recurse = recurse;
int save_xfer_dirs = xfer_dirs;
int save_prune_empty_dirs = prune_empty_dirs;
int senddir_fd = flags & GDL_IGNORE_FILTER_RULES ? -2 : -1;
if (dlen < 0) {
dlen = strlcpy(dirbuf, dirname, MAXPATHLEN);
if (dlen >= MAXPATHLEN)
return NULL;
dirname = dirbuf;
}
dirlist = flist_new(FLIST_TEMP, "get_dirlist");
recurse = 0;
xfer_dirs = 1;
send_directory(senddir_fd, dirlist, dirname, dlen, FLAG_CONTENT_DIR);
xfer_dirs = save_xfer_dirs;
recurse = save_recurse;
if (INFO_GTE(PROGRESS, 1))
flist_count_offset += dirlist->used;
prune_empty_dirs = 0;
dirlist->sorted = dirlist->files;
flist_sort_and_clean(dirlist, 0);
prune_empty_dirs = save_prune_empty_dirs;
if (DEBUG_GTE(FLIST, 3))
output_flist(dirlist);
return dirlist;
}