X-Git-Url: https://mattmccutchen.net/rsync/rsync.git/blobdiff_plain/5f78da202569ff1a4126c85a497156f5a7604e9a..9935066b704bcf2e6e48dac85cb1b4047d8f439d:/flist.c diff --git a/flist.c b/flist.c index ab729c80..19931a5f 100644 --- a/flist.c +++ b/flist.c @@ -1,18 +1,18 @@ -/* +/* Copyright (C) Andrew Tridgell 1996 Copyright (C) Paul Mackerras 1996 Copyright (C) 2001, 2002 by Martin Pool - + 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. @@ -21,10 +21,6 @@ /** @file flist.c * Generate and receive file lists * - * @todo Get rid of the string_area optimization. Efficiently - * allocating blocks is the responsibility of the system's malloc - * library, not of rsync. - * * @sa http://lists.samba.org/pipermail/rsync/2000-June/002351.html * **/ @@ -34,12 +30,20 @@ extern struct stats stats; extern int verbose; +extern int do_progress; +extern int am_root; extern int am_server; extern int always_checksum; +extern int module_id; +extern int ignore_errors; +extern int numeric_ids; extern int cvs_exclude; extern int recurse; +extern char curr_dir[MAXPATHLEN]; +extern char *files_from; +extern int filesfrom_fd; extern int one_file_system; extern int make_backups; @@ -51,110 +55,79 @@ extern int preserve_uid; extern int preserve_gid; extern int preserve_times; extern int relative_paths; +extern int implied_dirs; extern int copy_links; extern int copy_unsafe_links; -extern int remote_version; -extern int io_error; +extern int protocol_version; extern int sanitize_paths; extern int read_batch; extern int write_batch; -static char topsrcname[MAXPATHLEN]; +extern struct exclude_struct **exclude_list; +extern struct exclude_struct **server_exclude_list; +extern struct exclude_struct **local_exclude_list; -static struct exclude_struct **local_exclude_list; +int io_error; -static struct file_struct null_file; +static char empty_sum[MD4_SUM_LENGTH]; +static unsigned int file_struct_len; -static void clean_flist(struct file_list *flist, int strip_root); +static void clean_flist(struct file_list *flist, int strip_root, int no_dups); +static void output_flist(struct file_list *flist); - -static int show_build_progress_p(void) +void init_flist(void) { - extern int do_progress; + struct file_struct f; - return do_progress && verbose && recurse && !am_server; + /* Figure out how big the file_struct is without trailing padding */ + file_struct_len = ((char*)&f.flags - (char*)&f) + sizeof f.flags; } -/** - * True if we're local, etc, and should emit progress emssages. - **/ -static void emit_build_progress(const struct file_list *flist) + +static int show_filelist_p(void) { - rprintf(FINFO, " %d files...\r", flist->count); + return verbose && (recurse || files_from) && !am_server; } - -static void finish_build_progress(const struct file_list *flist) +static void start_filelist_progress(char *kind) { - if (verbose && recurse && !am_server) { - /* This overwrites the progress line, if any. */ - rprintf(FINFO, RSYNC_NAME ": %d files to consider.\n", - flist->count); - } + rprintf(FINFO, "%s ... ", kind); + if ((verbose > 1) || do_progress) + rprintf(FINFO, "\n"); + rflush(FINFO); } -void show_flist_stats(void) +static void emit_filelist_progress(const struct file_list *flist) { - /* Nothing yet */ + rprintf(FINFO, " %d files...\r", flist->count); } -static struct string_area *string_area_new(int size) +static void maybe_emit_filelist_progress(const struct file_list *flist) { - struct string_area *a; - - if (size <= 0) - size = ARENA_SIZE; - a = malloc(sizeof(*a)); - if (!a) - out_of_memory("string_area_new"); - a->current = a->base = malloc(size); - if (!a->current) - out_of_memory("string_area_new buffer"); - a->end = a->base + size; - a->next = NULL; - - return a; + if (do_progress && show_filelist_p() && ((flist->count % 100) == 0)) + emit_filelist_progress(flist); } -static void string_area_free(struct string_area *a) -{ - struct string_area *next; - - for (; a; a = next) { - next = a->next; - free(a->base); - } -} -static char *string_area_malloc(struct string_area **ap, int size) +static void finish_filelist_progress(const struct file_list *flist) { - char *p; - struct string_area *a; - - /* does the request fit into the current space? */ - a = *ap; - if (a->current + size >= a->end) { - /* no; get space, move new string_area to front of the list */ - a = string_area_new(size > ARENA_SIZE ? size : ARENA_SIZE); - a->next = *ap; - *ap = a; - } - - /* have space; do the "allocation." */ - p = a->current; - a->current += size; - return p; + if (do_progress) { + /* This overwrites the progress line */ + rprintf(FINFO, "%d file%sto consider\n", + flist->count, flist->count == 1 ? " " : "s "); + } else + rprintf(FINFO, "done\n"); } -static char *string_area_strdup(struct string_area **ap, const char *src) +void show_flist_stats(void) { - char *dest = string_area_malloc(ap, strlen(src) + 1); - return strcpy(dest, src); + /* Nothing yet */ } + static void list_file_entry(struct file_struct *f) { char perms[11]; @@ -165,76 +138,104 @@ static void list_file_entry(struct file_struct *f) permstring(perms, f->mode); +#if SUPPORT_LINKS if (preserve_links && S_ISLNK(f->mode)) { rprintf(FINFO, "%s %11.0f %s %s -> %s\n", perms, (double) f->length, timestring(f->modtime), - f_name(f), f->link); - } else { + f_name(f), f->u.link); + } else +#endif rprintf(FINFO, "%s %11.0f %s %s\n", perms, (double) f->length, timestring(f->modtime), f_name(f)); - } } -int readlink_stat(const char *Path, STRUCT_STAT * Buffer, char *Linkbuf) +/** + * Stat either a symlink or its referent, depending on the settings of + * copy_links, copy_unsafe_links, etc. + * + * @retval -1 on error + * + * @retval 0 for success + * + * @post If @p path is a symlink, then @p linkbuf (of size @c + * MAXPATHLEN) contains the symlink target. + * + * @post @p buffer contains information about the link or the + * referrent as appropriate, if they exist. + **/ +int readlink_stat(const char *path, STRUCT_STAT *buffer, char *linkbuf) { #if SUPPORT_LINKS - if (copy_links) { - return do_stat(Path, Buffer); - } - if (do_lstat(Path, Buffer) == -1) { + if (copy_links) + return do_stat(path, buffer); + if (do_lstat(path, buffer) == -1) return -1; - } - if (S_ISLNK(Buffer->st_mode)) { - int l; - if ((l = - readlink((char *) Path, Linkbuf, - MAXPATHLEN - 1)) == -1) { + if (S_ISLNK(buffer->st_mode)) { + int l = readlink((char *) path, linkbuf, MAXPATHLEN - 1); + if (l == -1) return -1; - } - Linkbuf[l] = 0; - if (copy_unsafe_links && (topsrcname[0] != '\0') && - unsafe_symlink(Linkbuf, topsrcname)) { - return do_stat(Path, Buffer); + linkbuf[l] = 0; + if (copy_unsafe_links && unsafe_symlink(linkbuf, path)) { + if (verbose > 1) { + rprintf(FINFO,"copying unsafe symlink \"%s\" -> \"%s\"\n", + path, linkbuf); + } + return do_stat(path, buffer); } } return 0; #else - return do_stat(Path, Buffer); + return do_stat(path, buffer); #endif } -int link_stat(const char *Path, STRUCT_STAT * Buffer) +int link_stat(const char *path, STRUCT_STAT * buffer) { #if SUPPORT_LINKS - if (copy_links) { - return do_stat(Path, Buffer); - } else { - return do_lstat(Path, Buffer); - } + if (copy_links) + return do_stat(path, buffer); + return do_lstat(path, buffer); #else - return do_stat(Path, Buffer); + return do_stat(path, buffer); #endif } /* - 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 + * 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 + * exclude_level is set to either SERVER_EXCLUDES or ALL_EXCLUDES. */ -static int check_exclude_file(int f, char *fname, STRUCT_STAT * st) +static int check_exclude_file(char *fname, int is_dir, int exclude_level) { - extern int delete_excluded; - - /* f is set to -1 when calculating deletion file list */ - if ((f == -1) && delete_excluded) { +#if 0 /* This currently never happens, so avoid a useless compare. */ + if (exclude_level == NO_EXCLUDES) return 0; +#endif + if (fname) { + /* never exclude '.', even if somebody does --exclude '*' */ + if (fname[0] == '.' && !fname[1]) + return 0; + /* Handle the -R version of the '.' dir. */ + if (fname[0] == '/') { + int len = strlen(fname); + if (fname[len-1] == '.' && fname[len-2] == '/') + return 0; + } } - if (check_exclude(fname, local_exclude_list, st)) { + if (server_exclude_list + && check_exclude(server_exclude_list, fname, is_dir)) + return 1; + if (exclude_level != ALL_EXCLUDES) + return 0; + if (exclude_list && check_exclude(exclude_list, fname, is_dir)) + return 1; + if (local_exclude_list + && check_exclude(local_exclude_list, fname, is_dir)) return 1; - } return 0; } @@ -252,17 +253,17 @@ static void set_filesystem(char *fname) static int to_wire_mode(mode_t mode) { - if (S_ISLNK(mode) && (_S_IFLNK != 0120000)) { +#if SUPPORT_LINKS + if (S_ISLNK(mode) && (_S_IFLNK != 0120000)) return (mode & ~(_S_IFMT)) | 0120000; - } +#endif return (int) mode; } static mode_t from_wire_mode(int mode) { - if ((mode & (_S_IFMT)) == 0120000 && (_S_IFLNK != 0120000)) { + if ((mode & (_S_IFMT)) == 0120000 && (_S_IFLNK != 0120000)) return (mode & ~(_S_IFMT)) | _S_IFLNK; - } return (mode_t) mode; } @@ -270,52 +271,65 @@ static mode_t from_wire_mode(int mode) static void send_directory(int f, struct file_list *flist, char *dir); static char *flist_dir; +static int flist_dir_len; /** * Make sure @p flist is big enough to hold at least @p flist->count * entries. **/ -static void flist_expand(struct file_list *flist) +void flist_expand(struct file_list *flist) { - if (flist->count >= flist->malloced) { - size_t new_bytes; - void *new_ptr; - - if (flist->malloced < 1000) - flist->malloced += 1000; - else - flist->malloced *= 2; + void *new_ptr; - new_bytes = sizeof(flist->files[0]) * flist->malloced; - - new_ptr = realloc(flist->files, new_bytes); + if (flist->count < flist->malloced) + return; - if (verbose >= 2) { - rprintf(FINFO, RSYNC_NAME ": expand file_list to %.0f bytes, did%s move\n", - (double) new_bytes, - (new_ptr == flist->files) ? " not" : ""); - } - - flist->files = (struct file_struct **) new_ptr; + 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->count) + flist->malloced = flist->count; + + if (flist->files) { + new_ptr = realloc_array(flist->files, + struct file_struct *, flist->malloced); + } else { + new_ptr = new_array(struct file_struct *, flist->malloced); + } - if (!flist->files) - out_of_memory("flist_expand"); + if (verbose >= 2) { + rprintf(FINFO, "[%s] expand file_list to %.0f bytes, did%s move\n", + who_am_i(), + (double) sizeof flist->files[0] * flist->malloced, + (new_ptr == flist->files) ? " not" : ""); } -} + flist->files = (struct file_struct **) new_ptr; + + if (!flist->files) + out_of_memory("flist_expand"); +} -static void send_file_entry(struct file_struct *file, int f, - unsigned base_flags) +void send_file_entry(struct file_struct *file, int f, unsigned short base_flags) { - unsigned char flags; - static time_t last_time; - static mode_t last_mode; - static DEV64_T last_rdev; - static uid_t last_uid; - static gid_t last_gid; + unsigned short flags; + static time_t modtime; + static mode_t mode; + static DEV64_T rdev, rdev_high; + static DEV64_T dev; + static uid_t uid; + static gid_t gid; static char lastname[MAXPATHLEN]; - char *fname; + char *fname, fbuf[MAXPATHLEN]; int l1, l2; if (f == -1) @@ -323,136 +337,209 @@ static void send_file_entry(struct file_struct *file, int f, if (!file) { write_byte(f, 0); + modtime = 0, mode = 0; + rdev = 0, rdev_high = 0, dev = 0; + uid = 0, gid = 0; + *lastname = '\0'; return; } - fname = f_name(file); + io_write_phase = "send_file_entry"; + + fname = f_name_to(file, fbuf); flags = base_flags; - if (file->mode == last_mode) - flags |= SAME_MODE; - if (file->rdev == last_rdev) - flags |= SAME_RDEV; - if (file->uid == last_uid) - flags |= SAME_UID; - if (file->gid == last_gid) - flags |= SAME_GID; - if (file->modtime == last_time) - flags |= SAME_TIME; + if (file->mode == mode) + flags |= XMIT_SAME_MODE; + else + mode = file->mode; + if (preserve_devices) { + if (protocol_version < 28) { + if (IS_DEVICE(mode)) { + if (file->u.rdev == rdev) { + /* Set both flags to simplify the test + * when writing the data. */ + flags |= XMIT_SAME_RDEV_pre28 + | XMIT_SAME_HIGH_RDEV; + } else + rdev = file->u.rdev; + } else + rdev = 0; + } else if (IS_DEVICE(mode)) { + if ((file->u.rdev & ~0xFF) == rdev_high) + flags |= XMIT_SAME_HIGH_RDEV; + else { + rdev = file->u.rdev; + rdev_high = rdev & ~0xFF; + } + } + } + if (file->uid == uid) + flags |= XMIT_SAME_UID; + else + uid = file->uid; + if (file->gid == gid) + flags |= XMIT_SAME_GID; + else + gid = file->gid; + if (file->modtime == modtime) + flags |= XMIT_SAME_TIME; + else + modtime = file->modtime; + +#if SUPPORT_HARD_LINKS + if (file->link_u.idev) { + if (file->F_DEV == dev) { + if (protocol_version >= 28) + flags |= XMIT_SAME_DEV; + } else + dev = file->F_DEV; + flags |= XMIT_HAS_IDEV_DATA; + } +#endif for (l1 = 0; lastname[l1] && (fname[l1] == lastname[l1]) && (l1 < 255); - l1++); - l2 = strlen(fname) - l1; + l1++) {} + l2 = strlen(fname+l1); if (l1 > 0) - flags |= SAME_NAME; + flags |= XMIT_SAME_NAME; if (l2 > 255) - flags |= LONG_NAME; - - /* we must make sure we don't send a zero flags byte or the other - end will terminate the flist transfer */ - if (flags == 0 && !S_ISDIR(file->mode)) - flags |= FLAG_DELETE; - if (flags == 0) - flags |= LONG_NAME; - - write_byte(f, flags); - if (flags & SAME_NAME) + flags |= 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 (!flags && !S_ISDIR(mode)) + flags |= XMIT_TOP_DIR; + if ((flags & 0xFF00) || !flags) { + flags |= XMIT_EXTENDED_FLAGS; + write_byte(f, flags); + write_byte(f, flags >> 8); + } else + write_byte(f, flags); + } else { + if (!(flags & 0xFF) && !S_ISDIR(mode)) + flags |= XMIT_TOP_DIR; + if (!(flags & 0xFF)) + flags |= XMIT_LONG_NAME; + write_byte(f, flags); + } + if (flags & XMIT_SAME_NAME) write_byte(f, l1); - if (flags & LONG_NAME) + if (flags & XMIT_LONG_NAME) write_int(f, l2); else write_byte(f, l2); write_buf(f, fname + l1, l2); write_longint(f, file->length); - if (!(flags & SAME_TIME)) - write_int(f, (int) file->modtime); - if (!(flags & SAME_MODE)) - write_int(f, to_wire_mode(file->mode)); - if (preserve_uid && !(flags & SAME_UID)) { - add_uid(file->uid); - write_int(f, (int) file->uid); + if (!(flags & XMIT_SAME_TIME)) + write_int(f, modtime); + if (!(flags & XMIT_SAME_MODE)) + write_int(f, to_wire_mode(mode)); + if (preserve_uid && !(flags & XMIT_SAME_UID)) { + if (!numeric_ids) + add_uid(uid); + write_int(f, uid); } - if (preserve_gid && !(flags & SAME_GID)) { - add_gid(file->gid); - write_int(f, (int) file->gid); + if (preserve_gid && !(flags & XMIT_SAME_GID)) { + if (!numeric_ids) + add_gid(gid); + write_int(f, gid); + } + if (preserve_devices && IS_DEVICE(mode)) { + /* If XMIT_SAME_HIGH_RDEV is off, XMIT_SAME_RDEV_pre28 is + * also off. */ + if (!(flags & XMIT_SAME_HIGH_RDEV)) + write_int(f, rdev); + else if (protocol_version >= 28) + write_byte(f, rdev); } - if (preserve_devices && IS_DEVICE(file->mode) - && !(flags & SAME_RDEV)) - write_int(f, (int) file->rdev); #if SUPPORT_LINKS - if (preserve_links && S_ISLNK(file->mode)) { - write_int(f, strlen(file->link)); - write_buf(f, file->link, strlen(file->link)); + if (preserve_links && S_ISLNK(mode)) { + int len = strlen(file->u.link); + write_int(f, len); + write_buf(f, file->u.link, len); } #endif #if SUPPORT_HARD_LINKS - if (preserve_hard_links && S_ISREG(file->mode)) { - if (remote_version < 26) { + if (flags & XMIT_HAS_IDEV_DATA) { + if (protocol_version < 26) { /* 32-bit dev_t and ino_t */ - write_int(f, (int) file->dev); - write_int(f, (int) file->inode); + write_int(f, dev); + write_int(f, file->F_INODE); } else { /* 64-bit dev_t and ino_t */ - write_longint(f, file->dev); - write_longint(f, file->inode); + if (!(flags & XMIT_SAME_DEV)) + write_longint(f, dev); + write_longint(f, file->F_INODE); } } #endif if (always_checksum) { - if (remote_version < 21) { - write_buf(f, file->sum, 2); - } else { - write_buf(f, file->sum, MD4_SUM_LENGTH); + char *sum; + if (S_ISREG(mode)) + sum = file->u.sum; + else if (protocol_version < 28) { + /* Prior to 28, we sent a useless set of nulls. */ + sum = empty_sum; + } else + sum = NULL; + if (sum) { + write_buf(f, sum, protocol_version < 21? 2 + : MD4_SUM_LENGTH); } } - last_mode = file->mode; - last_rdev = file->rdev; - last_uid = file->uid; - last_gid = file->gid; - last_time = file->modtime; - strlcpy(lastname, fname, MAXPATHLEN); - lastname[MAXPATHLEN - 1] = 0; + + io_write_phase = "unknown"; } -static void receive_file_entry(struct file_struct **fptr, - unsigned flags, int f) +void receive_file_entry(struct file_struct **fptr, unsigned short flags, + struct file_list *flist, int f) { - static time_t last_time; - static mode_t last_mode; - static DEV64_T last_rdev; - static uid_t last_uid; - static gid_t last_gid; - static char lastname[MAXPATHLEN]; + static time_t modtime; + static mode_t mode; + static DEV64_T rdev, rdev_high; + static DEV64_T dev; + static uid_t uid; + static gid_t gid; + static char lastname[MAXPATHLEN], *lastdir; + static int lastdir_len = -1; char thisname[MAXPATHLEN]; unsigned int l1 = 0, l2 = 0; - char *p; + int alloc_len, basename_len, dirname_len, linkname_len, sum_len; + OFF_T file_length; + char *basename, *dirname, *bp; struct file_struct *file; - if (flags & SAME_NAME) + if (!fptr) { + modtime = 0, mode = 0; + rdev = 0, rdev_high = 0, dev = 0; + uid = 0, gid = 0; + *lastname = '\0'; + return; + } + + if (flags & XMIT_SAME_NAME) l1 = read_byte(f); - if (flags & LONG_NAME) + if (flags & XMIT_LONG_NAME) l2 = read_int(f); else l2 = read_byte(f); - file = (struct file_struct *) malloc(sizeof(*file)); - if (!file) - out_of_memory("receive_file_entry"); - memset((char *) file, 0, sizeof(*file)); - (*fptr) = file; - if (l2 >= MAXPATHLEN - l1) { rprintf(FERROR, "overflow: flags=0x%x l1=%d l2=%d lastname=%s\n", @@ -465,258 +552,341 @@ static void receive_file_entry(struct file_struct **fptr, thisname[l1 + l2] = 0; strlcpy(lastname, thisname, MAXPATHLEN); - lastname[MAXPATHLEN - 1] = 0; clean_fname(thisname); - if (sanitize_paths) { + if (sanitize_paths) sanitize_path(thisname, NULL); + + if ((basename = strrchr(thisname, '/')) != NULL) { + dirname_len = ++basename - thisname; /* counts future '\0' */ + if (lastdir_len == dirname_len - 1 + && strncmp(thisname, lastdir, lastdir_len) == 0) { + dirname = lastdir; + dirname_len = 0; /* indicates no copy is needed */ + } else + dirname = thisname; + } else { + basename = thisname; + dirname = NULL; + dirname_len = 0; + } + basename_len = strlen(basename) + 1; /* count the '\0' */ + + file_length = read_longint(f); + if (!(flags & XMIT_SAME_TIME)) + modtime = (time_t)read_int(f); + if (!(flags & XMIT_SAME_MODE)) + mode = from_wire_mode(read_int(f)); + + if (preserve_uid && !(flags & XMIT_SAME_UID)) + uid = (uid_t)read_int(f); + if (preserve_gid && !(flags & XMIT_SAME_GID)) + gid = (gid_t)read_int(f); + + if (preserve_devices) { + if (protocol_version < 28) { + if (IS_DEVICE(mode)) { + if (!(flags & XMIT_SAME_RDEV_pre28)) + rdev = (DEV64_T)read_int(f); + } else + rdev = 0; + } else if (IS_DEVICE(mode)) { + if (!(flags & XMIT_SAME_HIGH_RDEV)) { + rdev = (DEV64_T)read_int(f); + rdev_high = rdev & ~0xFF; + } else + rdev = rdev_high | (DEV64_T)read_byte(f); + } } - if ((p = strrchr(thisname, '/'))) { - static char *lastdir; - *p = 0; - if (lastdir && strcmp(thisname, lastdir) == 0) { - file->dirname = lastdir; - } else { - file->dirname = strdup(thisname); - lastdir = file->dirname; +#if SUPPORT_LINKS + if (preserve_links && S_ISLNK(mode)) { + linkname_len = read_int(f) + 1; /* count the '\0' */ + if (linkname_len <= 0 || linkname_len > MAXPATHLEN) { + rprintf(FERROR, "overflow: linkname_len=%d\n", + linkname_len - 1); + overflow("receive_file_entry"); } - file->basename = strdup(p + 1); - } else { - file->dirname = NULL; - file->basename = strdup(thisname); } + else +#endif + linkname_len = 0; - if (!file->basename) - out_of_memory("receive_file_entry 1"); + sum_len = always_checksum && S_ISREG(mode) ? MD4_SUM_LENGTH : 0; + alloc_len = file_struct_len + dirname_len + basename_len + + linkname_len + sum_len; + bp = pool_alloc(flist->file_pool, alloc_len, "receive_file_entry"); - file->flags = flags; - file->length = read_longint(f); - file->modtime = - (flags & SAME_TIME) ? last_time : (time_t) read_int(f); - file->mode = - (flags & SAME_MODE) ? last_mode : from_wire_mode(read_int(f)); - if (preserve_uid) - file->uid = - (flags & SAME_UID) ? last_uid : (uid_t) read_int(f); - if (preserve_gid) - file->gid = - (flags & SAME_GID) ? last_gid : (gid_t) read_int(f); - if (preserve_devices && IS_DEVICE(file->mode)) - file->rdev = - (flags & SAME_RDEV) ? last_rdev : (dev_t) read_int(f); - - if (preserve_links && S_ISLNK(file->mode)) { - int l = read_int(f); - if (l < 0) { - rprintf(FERROR, "overflow: l=%d\n", l); - overflow("receive_file_entry"); - } - file->link = (char *) malloc(l + 1); - if (!file->link) - out_of_memory("receive_file_entry 2"); - read_sbuf(f, file->link, l); - if (sanitize_paths) { - sanitize_path(file->link, file->dirname); - } + file = *fptr = (struct file_struct *)bp; + memset(bp, 0, file_struct_len); + bp += file_struct_len; + + file->flags = flags & XMIT_TOP_DIR ? FLAG_TOP_DIR : 0; + file->modtime = modtime; + file->length = file_length; + file->mode = mode; + file->uid = uid; + file->gid = gid; + + if (dirname_len) { + file->dirname = lastdir = bp; + lastdir_len = dirname_len - 1; + memcpy(bp, dirname, dirname_len - 1); + bp += dirname_len; + bp[-1] = '\0'; + } else if (dirname) + file->dirname = dirname; + + file->basename = bp; + memcpy(bp, basename, basename_len); + bp += basename_len; + + if (preserve_devices && IS_DEVICE(mode)) + file->u.rdev = rdev; + +#if SUPPORT_LINKS + if (linkname_len) { + file->u.link = bp; + read_sbuf(f, bp, linkname_len - 1); + if (sanitize_paths) + sanitize_path(bp, lastdir); + bp += linkname_len; } +#endif + #if SUPPORT_HARD_LINKS - if (preserve_hard_links && S_ISREG(file->mode)) { - if (remote_version < 26) { - file->dev = read_int(f); - file->inode = read_int(f); + if (preserve_hard_links && protocol_version < 28 && S_ISREG(mode)) + flags |= XMIT_HAS_IDEV_DATA; + if (flags & XMIT_HAS_IDEV_DATA && flist->hlink_pool) { + INO64_T inode; + file->link_u.idev = pool_talloc(flist->hlink_pool, + struct idev, 1, "inode_table"); + if (protocol_version < 26) { + dev = read_int(f); + inode = read_int(f); } else { - file->dev = read_longint(f); - file->inode = read_longint(f); + if (!(flags & XMIT_SAME_DEV)) + dev = read_longint(f); + inode = read_longint(f); + } + if (flist->hlink_pool) { + file->F_INODE = inode; + file->F_DEV = dev; } } #endif if (always_checksum) { - file->sum = (char *) malloc(MD4_SUM_LENGTH); - if (!file->sum) - out_of_memory("md4 sum"); - if (remote_version < 21) { - read_buf(f, file->sum, 2); - } else { - read_buf(f, file->sum, MD4_SUM_LENGTH); + char *sum; + if (sum_len) { + file->u.sum = sum = bp; + /*bp += sum_len;*/ + } else if (protocol_version < 28) { + /* Prior to 28, we get a useless set of nulls. */ + sum = empty_sum; + } else + sum = NULL; + if (sum) { + read_buf(f, sum, protocol_version < 21? 2 + : MD4_SUM_LENGTH); } } - last_mode = file->mode; - last_rdev = file->rdev; - last_uid = file->uid; - last_gid = file->gid; - last_time = file->modtime; - if (!preserve_perms) { extern int orig_umask; /* set an appropriate set of permissions based on original - permissions and umask. This emulates what GNU cp does */ + * permissions and umask. This emulates what GNU cp does */ file->mode &= ~orig_umask; } } -/* determine if a file in a different filesstem should be skipped - when one_file_system is set. We bascally only want to include - the mount points - but they can be hard to find! */ -static int skip_filesystem(char *fname, STRUCT_STAT * st) -{ - STRUCT_STAT st2; - char *p = strrchr(fname, '/'); - - /* skip all but directories */ - if (!S_ISDIR(st->st_mode)) - return 1; - - /* if its not a subdirectory then allow */ - if (!p) - return 0; - - *p = 0; - if (link_stat(fname, &st2)) { - *p = '/'; - return 0; - } - *p = '/'; - - return (st2.st_dev != filesystem_dev); -} - -#define STRDUP(ap, p) (ap ? string_area_strdup(ap, p) : strdup(p)) -/* IRIX cc cares that the operands to the ternary have the same type. */ -#define MALLOC(ap, i) (ap ? (void*) string_area_malloc(ap, i) : malloc(i)) - -/* create a file_struct for a named file */ -struct file_struct *make_file(int f, char *fname, struct string_area **ap, - int noexcludes) +/** + * Create a file_struct for a named file by reading its stat() + * information and performing extensive checks against global + * options. + * + * @return the new file, or NULL if there was an error or this file + * should be excluded. + * + * @todo There is a small optimization opportunity here to avoid + * stat()ing the file in some circumstances, which has a certain cost. + * We are called immediately after doing readdir(), and so we may + * already know the d_type of the file. We could for example avoid + * statting directories if we're not recursing, but this is not a very + * important case. Some systems may not have d_type. + **/ +struct file_struct *make_file(char *fname, + struct file_list *flist, int exclude_level) { + static char *lastdir; + static int lastdir_len = -1; struct file_struct *file; STRUCT_STAT st; char sum[SUM_LENGTH]; - char *p; - char cleaned_name[MAXPATHLEN]; - char linkbuf[MAXPATHLEN]; - extern int module_id; - - strlcpy(cleaned_name, fname, MAXPATHLEN); - cleaned_name[MAXPATHLEN - 1] = 0; - clean_fname(cleaned_name); - if (sanitize_paths) { - sanitize_path(cleaned_name, NULL); + char thisname[MAXPATHLEN]; + char linkname[MAXPATHLEN]; + int alloc_len, basename_len, dirname_len, linkname_len, sum_len; + char *basename, *dirname, *bp; + unsigned short flags = 0; + + + if (strlcpy(thisname, fname, sizeof thisname) + >= sizeof thisname - flist_dir_len) { + rprintf(FINFO, "skipping overly long name: %s\n", fname); + return NULL; } - fname = cleaned_name; + clean_fname(thisname); + if (sanitize_paths) + sanitize_path(thisname, NULL); memset(sum, 0, SUM_LENGTH); - if (readlink_stat(fname, &st, linkbuf) != 0) { + if (readlink_stat(thisname, &st, linkname) != 0) { int save_errno = errno; - if ((errno == ENOENT) && copy_links && !noexcludes) { - /* symlink pointing nowhere, see if excluded */ - memset((char *) &st, 0, sizeof(st)); - if (check_exclude_file(f, fname, &st)) { + if (errno == ENOENT && exclude_level != NO_EXCLUDES) { + /* either symlink pointing nowhere or file that + * was removed during rsync run; see if excluded + * before reporting an error */ + if (check_exclude_file(thisname, 0, exclude_level)) { /* file is excluded anyway, ignore silently */ return NULL; } } - io_error = 1; - rprintf(FERROR, "readlink %s: %s\n", - fname, strerror(save_errno)); + io_error |= IOERR_GENERAL; + rprintf(FERROR, "readlink %s failed: %s\n", + full_fname(thisname), strerror(save_errno)); return NULL; } - /* we use noexcludes from backup.c */ - if (noexcludes) + /* backup.c calls us with exclude_level set to NO_EXCLUDES. */ + if (exclude_level == NO_EXCLUDES) goto skip_excludes; - if (S_ISDIR(st.st_mode) && !recurse) { - rprintf(FINFO, "skipping directory %s\n", fname); + if (S_ISDIR(st.st_mode) && !recurse && !files_from) { + rprintf(FINFO, "skipping directory %s\n", thisname); return NULL; } - if (one_file_system && st.st_dev != filesystem_dev) { - if (skip_filesystem(fname, &st)) - return NULL; - } + /* We only care about directories 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 + && S_ISDIR(st.st_mode)) + flags |= FLAG_MOUNT_POINT; - if (check_exclude_file(f, fname, &st)) + if (check_exclude_file(thisname, S_ISDIR(st.st_mode) != 0, exclude_level)) return NULL; - - if (lp_ignore_nonreadable(module_id) && access(fname, R_OK) != 0) + if (lp_ignore_nonreadable(module_id) && access(thisname, R_OK) != 0) return NULL; skip_excludes: - if (verbose > 2) - rprintf(FINFO, "make_file(%d,%s)\n", f, fname); + if (verbose > 2) { + rprintf(FINFO, "[%s] make_file(%s,*,%d)\n", + who_am_i(), thisname, exclude_level); + } - file = (struct file_struct *) malloc(sizeof(*file)); - if (!file) - out_of_memory("make_file"); - memset((char *) file, 0, sizeof(*file)); - - if ((p = strrchr(fname, '/'))) { - static char *lastdir; - *p = 0; - if (lastdir && strcmp(fname, lastdir) == 0) { - file->dirname = lastdir; - } else { - file->dirname = strdup(fname); - lastdir = file->dirname; - } - file->basename = STRDUP(ap, p + 1); - *p = '/'; + if ((basename = strrchr(thisname, '/')) != NULL) { + dirname_len = ++basename - thisname; /* counts future '\0' */ + if (lastdir_len == dirname_len - 1 + && strncmp(thisname, lastdir, lastdir_len) == 0) { + dirname = lastdir; + dirname_len = 0; /* indicates no copy is needed */ + } else + dirname = thisname; + } else { + basename = thisname; + dirname = NULL; + dirname_len = 0; + } + basename_len = strlen(basename) + 1; /* count the '\0' */ + +#if SUPPORT_LINKS + linkname_len = S_ISLNK(st.st_mode) ? strlen(linkname) + 1 : 0; +#else + linkname_len = 0; +#endif + + sum_len = always_checksum && S_ISREG(st.st_mode) ? MD4_SUM_LENGTH : 0; + + alloc_len = file_struct_len + dirname_len + basename_len + + linkname_len + sum_len; + if (flist) { + bp = pool_alloc(flist->file_pool, alloc_len, + "receive_file_entry"); } else { - file->dirname = NULL; - file->basename = STRDUP(ap, fname); + if (!(bp = new_array(char, alloc_len))) + out_of_memory("receive_file_entry"); } + file = (struct file_struct *)bp; + memset(bp, 0, file_struct_len); + bp += file_struct_len; + + file->flags = flags; file->modtime = st.st_mtime; file->length = st.st_size; file->mode = st.st_mode; file->uid = st.st_uid; file->gid = st.st_gid; - file->dev = st.st_dev; - file->inode = st.st_ino; + +#if SUPPORT_HARD_LINKS + if (flist && flist->hlink_pool) { + if (protocol_version < 28) { + if (S_ISREG(st.st_mode)) + file->link_u.idev = pool_talloc( + flist->hlink_pool, struct idev, 1, + "inode_table"); + } else { + if (!S_ISDIR(st.st_mode) && st.st_nlink > 1) + file->link_u.idev = pool_talloc( + flist->hlink_pool, struct idev, 1, + "inode_table"); + } + } + if (file->link_u.idev) { + file->F_DEV = st.st_dev; + file->F_INODE = st.st_ino; + } +#endif + + if (dirname_len) { + file->dirname = lastdir = bp; + lastdir_len = dirname_len - 1; + memcpy(bp, dirname, dirname_len - 1); + bp += dirname_len; + bp[-1] = '\0'; + } else if (dirname) + file->dirname = dirname; + + file->basename = bp; + memcpy(bp, basename, basename_len); + bp += basename_len; + #ifdef HAVE_STRUCT_STAT_ST_RDEV - file->rdev = st.st_rdev; + if (preserve_devices && IS_DEVICE(st.st_mode)) + file->u.rdev = st.st_rdev; #endif #if SUPPORT_LINKS - if (S_ISLNK(st.st_mode)) { - file->link = STRDUP(ap, linkbuf); + if (linkname_len) { + file->u.link = bp; + memcpy(bp, linkname, linkname_len); + bp += linkname_len; } #endif - if (always_checksum) { - file->sum = (char *) MALLOC(ap, MD4_SUM_LENGTH); - if (!file->sum) - out_of_memory("md4 sum"); - /* drat. we have to provide a null checksum for non-regular - files in order to be compatible with earlier versions - of rsync */ - if (S_ISREG(st.st_mode)) { - file_checksum(fname, file->sum, st.st_size); - } else { - memset(file->sum, 0, MD4_SUM_LENGTH); - } + if (sum_len) { + file->u.sum = bp; + file_checksum(thisname, bp, st.st_size); + /*bp += sum_len;*/ } - if (flist_dir) { - static char *lastdir; - if (lastdir && strcmp(lastdir, flist_dir) == 0) { - file->basedir = lastdir; - } else { - file->basedir = strdup(flist_dir); - lastdir = file->basedir; - } - } else { - file->basedir = NULL; - } + file->basedir = flist_dir; if (!S_ISDIR(st.st_mode)) stats.total_size += st.st_size; @@ -725,167 +895,192 @@ struct file_struct *make_file(int f, char *fname, struct string_area **ap, } - void send_file_name(int f, struct file_list *flist, char *fname, - int recursive, unsigned base_flags) + int recursive, unsigned short base_flags) { struct file_struct *file; + char fbuf[MAXPATHLEN]; + extern int delete_excluded; - file = make_file(f, fname, &flist->string_area, 0); + /* f is set to -1 when calculating deletion file list */ + file = make_file(fname, flist, + f == -1 && delete_excluded? SERVER_EXCLUDES : ALL_EXCLUDES); if (!file) return; - if (show_build_progress_p() & !(flist->count % 100)) - emit_build_progress(flist); + maybe_emit_filelist_progress(flist); flist_expand(flist); - if (write_batch) /* dw */ - file->flags = FLAG_DELETE; + if (write_batch) + file->flags |= FLAG_TOP_DIR; - if (strcmp(file->basename, "")) { + if (file->basename[0]) { flist->files[flist->count++] = file; send_file_entry(file, f, base_flags); } - if (S_ISDIR(file->mode) && recursive) { - struct exclude_struct **last_exclude_list = - local_exclude_list; - send_directory(f, flist, f_name(file)); + if (recursive && S_ISDIR(file->mode) + && !(file->flags & FLAG_MOUNT_POINT)) { + struct exclude_struct **last_exclude_list = local_exclude_list; + send_directory(f, flist, f_name_to(file, fbuf)); local_exclude_list = last_exclude_list; return; } } - static void send_directory(int f, struct file_list *flist, char *dir) { DIR *d; struct dirent *di; char fname[MAXPATHLEN]; - int l; + unsigned int offset; char *p; d = opendir(dir); if (!d) { - io_error = 1; - rprintf(FERROR, "opendir(%s): %s\n", dir, strerror(errno)); + io_error |= IOERR_GENERAL; + rprintf(FERROR, "opendir %s failed: %s\n", + full_fname(dir), strerror(errno)); return; } - strlcpy(fname, dir, MAXPATHLEN); - l = strlen(fname); - if (fname[l - 1] != '/') { - if (l == MAXPATHLEN - 1) { - io_error = 1; - rprintf(FERROR, - "skipping long-named directory %s\n", - fname); + offset = strlcpy(fname, dir, MAXPATHLEN); + p = fname + offset; + if (offset >= MAXPATHLEN || p[-1] != '/') { + if (offset >= MAXPATHLEN - 1) { + io_error |= IOERR_GENERAL; + rprintf(FERROR, "skipping long-named directory: %s\n", + full_fname(fname)); closedir(d); return; } - strlcat(fname, "/", MAXPATHLEN); - l++; + *p++ = '/'; + offset++; } - p = fname + strlen(fname); local_exclude_list = NULL; if (cvs_exclude) { - if (strlen(fname) + strlen(".cvsignore") <= MAXPATHLEN - 1) { - strcpy(p, ".cvsignore"); - local_exclude_list = - make_exclude_list(fname, NULL, 0, 0); - } else { - io_error = 1; + if (strlcpy(p, ".cvsignore", MAXPATHLEN - offset) + < MAXPATHLEN - offset) + add_exclude_file(&local_exclude_list,fname,MISSING_OK,ADD_EXCLUDE); + else { + io_error |= IOERR_GENERAL; rprintf(FINFO, "cannot cvs-exclude in long-named directory %s\n", - fname); + full_fname(fname)); } } - for (di = readdir(d); di; di = readdir(d)) { + for (errno = 0, di = readdir(d); di; errno = 0, di = readdir(d)) { char *dname = d_name(di); - if (strcmp(dname, ".") == 0 || strcmp(dname, "..") == 0) + if (dname[0] == '.' && (dname[1] == '\0' + || (dname[1] == '.' && dname[2] == '\0'))) continue; - strlcpy(p, dname, MAXPATHLEN - l); - send_file_name(f, flist, fname, recurse, 0); + if (strlcpy(p, dname, MAXPATHLEN - offset) < MAXPATHLEN - offset) + send_file_name(f, flist, fname, recurse, 0); + else { + io_error |= IOERR_GENERAL; + rprintf(FINFO, + "cannot send long-named file %s\n", + full_fname(fname)); + } } - - if (local_exclude_list) { - add_exclude_list("!", &local_exclude_list, 0); + if (errno) { + io_error |= IOERR_GENERAL; + rprintf(FERROR, "readdir(%s): (%d) %s\n", + dir, errno, strerror(errno)); } + if (local_exclude_list) + free_exclude_list(&local_exclude_list); /* Zeros pointer too */ + closedir(d); } -/* - * - * I *think* f==-1 means that the list should just be built in memory - * and not transmitted. But who can tell? -- mbp - */ +/** + * The delete_files() function in receiver.c sets f to -1 so that we just + * construct the file list in memory without sending it over the wire. It + * also has the side-effect of ignoring user-excludes if delete_excluded + * is set (so that the delete list includes user-excluded files). + **/ struct file_list *send_file_list(int f, int argc, char *argv[]) { - int i, l; + int l; STRUCT_STAT st; - char *p, *dir, *olddir; + char *p, *dir, olddir[sizeof curr_dir]; char lastpath[MAXPATHLEN] = ""; struct file_list *flist; int64 start_write; + int use_ff_fd = 0; - if (verbose && recurse && !am_server && f != -1) { - rprintf(FINFO, RSYNC_NAME ": building file list...\n"); - if (verbose > 1) - rprintf(FINFO, "\n"); - rflush(FINFO); - } + if (show_filelist_p() && f != -1) + start_filelist_progress("building file list"); start_write = stats.total_written; - flist = flist_new(); + flist = flist_new(f == -1 ? WITHOUT_HLINK : WITH_HLINK, + "send_file_list"); if (f != -1) { - io_start_buffering(f); + io_start_buffering_out(f); + if (filesfrom_fd >= 0) { + if (argv[0] && !push_dir(argv[0])) { + rprintf(FERROR, "push_dir %s failed: %s\n", + full_fname(argv[0]), strerror(errno)); + exit_cleanup(RERR_FILESELECT); + } + use_ff_fd = 1; + } } - for (i = 0; i < argc; i++) { - char *fname = topsrcname; + while (1) { + char fname2[MAXPATHLEN]; + char *fname = fname2; - strlcpy(fname, argv[i], MAXPATHLEN); + if (use_ff_fd) { + if (read_filesfrom_line(filesfrom_fd, fname) == 0) + break; + sanitize_path(fname, NULL); + } else { + if (argc-- == 0) + break; + strlcpy(fname, *argv++, MAXPATHLEN); + if (sanitize_paths) + sanitize_path(fname, NULL); + } l = strlen(fname); - if (l != 1 && fname[l - 1] == '/') { - if ((l == 2) && (fname[0] == '.')) { - /* Turn ./ into just . rather than ./. - This was put in to avoid a problem with - rsync -aR --delete from ./ - The send_file_name() below of ./ was - mysteriously preventing deletes */ - fname[1] = 0; - } else { - strlcat(fname, ".", MAXPATHLEN); + if (fname[l - 1] == '/') { + if (l == 2 && fname[0] == '.') { + /* Turn "./" into just "." rather than "./." */ + fname[1] = '\0'; + } else if (l < MAXPATHLEN) { + fname[l++] = '.'; + fname[l] = '\0'; } } if (link_stat(fname, &st) != 0) { if (f != -1) { - io_error = 1; - rprintf(FERROR, "link_stat %s : %s\n", - fname, strerror(errno)); + io_error |= IOERR_GENERAL; + rprintf(FERROR, "link_stat %s failed: %s\n", + full_fname(fname), strerror(errno)); } continue; } - if (S_ISDIR(st.st_mode) && !recurse) { + if (S_ISDIR(st.st_mode) && !recurse && !files_from) { rprintf(FINFO, "skipping directory %s\n", fname); continue; } dir = NULL; - olddir = NULL; + olddir[0] = '\0'; if (!relative_paths) { p = strrchr(fname, '/'); @@ -897,31 +1092,37 @@ struct file_list *send_file_list(int f, int argc, char *argv[]) dir = fname; fname = p + 1; } - } else if (f != -1 && (p = strrchr(fname, '/'))) { + } else if (f != -1 && implied_dirs && (p=strrchr(fname,'/')) && p != fname) { /* this ensures we send the intermediate directories, thus getting their permissions right */ + char *lp = lastpath, *fn = fname, *slash = fname; *p = 0; - if (strcmp(lastpath, fname)) { - strlcpy(lastpath, fname, sizeof(lastpath)); - *p = '/'; - for (p = fname + 1; (p = strchr(p, '/')); - p++) { - int copy_links_saved = copy_links; - int recurse_saved = recurse; - *p = 0; - copy_links = copy_unsafe_links; - /* set recurse to 1 to prevent make_file - from ignoring directory, but still - turn off the recursive parameter to - send_file_name */ - recurse = 1; - send_file_name(f, flist, fname, 0, - 0); - copy_links = copy_links_saved; - recurse = recurse_saved; - *p = '/'; + /* Skip any initial directories in our path that we + * have in common with lastpath. */ + while (*fn && *lp == *fn) { + if (*fn == '/') + slash = fn; + lp++, fn++; + } + *p = '/'; + if (fn != p || (*lp && *lp != '/')) { + int copy_links_saved = copy_links; + int recurse_saved = recurse; + copy_links = copy_unsafe_links; + /* set recurse to 1 to prevent make_file + * from ignoring directory, but still + * turn off the recursive parameter to + * send_file_name */ + recurse = 1; + while ((slash = strchr(slash+1, '/')) != 0) { + *slash = 0; + send_file_name(f, flist, fname, 0, 0); + *slash = '/'; } - } else { + copy_links = copy_links_saved; + recurse = recurse_saved; + *p = 0; + strlcpy(lastpath, fname, sizeof lastpath); *p = '/'; } } @@ -930,63 +1131,76 @@ struct file_list *send_file_list(int f, int argc, char *argv[]) fname = "."; if (dir && *dir) { - olddir = push_dir(dir, 1); + static char *lastdir; + static int lastdir_len; - if (!olddir) { - io_error = 1; - rprintf(FERROR, "push_dir %s : %s\n", - dir, strerror(errno)); + strcpy(olddir, curr_dir); /* can't overflow */ + + if (!push_dir(dir)) { + io_error |= IOERR_GENERAL; + rprintf(FERROR, "push_dir %s failed: %s\n", + full_fname(dir), strerror(errno)); continue; } - flist_dir = dir; + if (lastdir && strcmp(lastdir, dir) == 0) { + flist_dir = lastdir; + flist_dir_len = lastdir_len; + } else { + flist_dir = lastdir = strdup(dir); + flist_dir_len = lastdir_len = strlen(dir); + } } if (one_file_system) set_filesystem(fname); - send_file_name(f, flist, fname, recurse, FLAG_DELETE); + send_file_name(f, flist, fname, recurse, XMIT_TOP_DIR); - if (olddir != NULL) { + if (olddir[0]) { flist_dir = NULL; - if (pop_dir(olddir) != 0) { - rprintf(FERROR, "pop_dir %s : %s\n", - dir, strerror(errno)); + flist_dir_len = 0; + if (!pop_dir(olddir)) { + rprintf(FERROR, "pop_dir %s failed: %s\n", + full_fname(dir), strerror(errno)); exit_cleanup(RERR_FILESELECT); } } } - topsrcname[0] = '\0'; - if (f != -1) { send_file_entry(NULL, f, 0); + + if (show_filelist_p()) + finish_filelist_progress(flist); } - finish_build_progress(flist); + if (flist->hlink_pool) + { + pool_destroy(flist->hlink_pool); + flist->hlink_pool = NULL; + } - clean_flist(flist, 0); + clean_flist(flist, 0, 0); - /* now send the uid/gid list. This was introduced in protocol - version 15 */ - if (f != -1 && remote_version >= 15) { + if (f != -1) { + /* Now send the uid/gid list. This was introduced in + * protocol version 15 */ send_uid_list(f); - } - /* if protocol version is >= 17 then send the io_error flag */ - if (f != -1 && remote_version >= 17) { - extern int module_id; + /* send the io_error flag */ write_int(f, lp_ignore_errors(module_id) ? 0 : io_error); - } - if (f != -1) { - io_end_buffering(f); + io_end_buffering(); stats.flist_size = stats.total_written - start_write; stats.num_files = flist->count; - if (write_batch) /* dw */ + if (write_batch) write_batch_flist_info(flist->count, flist->files); } + if (verbose > 3) + output_flist(flist); + if (verbose > 2) rprintf(FINFO, "send_file_list done\n"); @@ -997,81 +1211,78 @@ struct file_list *send_file_list(int f, int argc, char *argv[]) struct file_list *recv_file_list(int f) { struct file_list *flist; - unsigned char flags; + unsigned short flags; int64 start_read; extern int list_only; - if (verbose && recurse && !am_server) { - rprintf(FINFO, "receiving file list ... "); - rflush(FINFO); - } + if (show_filelist_p()) + start_filelist_progress("receiving file list"); start_read = stats.total_read; - flist = (struct file_list *) malloc(sizeof(flist[0])); - if (!flist) - goto oom; + flist = flist_new(WITH_HLINK, "recv_file_list"); flist->count = 0; flist->malloced = 1000; - flist->files = - (struct file_struct **) malloc(sizeof(flist->files[0]) * - flist->malloced); + flist->files = new_array(struct file_struct *, flist->malloced); if (!flist->files) goto oom; - for (flags = read_byte(f); flags; flags = read_byte(f)) { + while ((flags = read_byte(f)) != 0) { int i = flist->count; - + flist_expand(flist); - receive_file_entry(&flist->files[i], flags, f); + if (protocol_version >= 28 && (flags & XMIT_EXTENDED_FLAGS)) + flags |= read_byte(f) << 8; + receive_file_entry(&flist->files[i], flags, flist, f); if (S_ISREG(flist->files[i]->mode)) stats.total_size += flist->files[i]->length; flist->count++; - if (verbose > 2) + maybe_emit_filelist_progress(flist); + + if (verbose > 2) { rprintf(FINFO, "recv_file_name(%s)\n", f_name(flist->files[i])); + } } - + receive_file_entry(NULL, 0, NULL, 0); /* Signal that we're done. */ if (verbose > 2) rprintf(FINFO, "received %d names\n", flist->count); - clean_flist(flist, relative_paths); + if (show_filelist_p()) + finish_filelist_progress(flist); - if (verbose && recurse && !am_server) { - rprintf(FINFO, "done\n"); - } + clean_flist(flist, relative_paths, 1); - /* now recv the uid/gid list. This was introduced in protocol version 15 */ - if (f != -1 && remote_version >= 15) { + if (f != -1) { + /* Now send the uid/gid list. This was introduced in + * protocol version 15 */ recv_uid_list(f, flist); - } - /* if protocol version is >= 17 then recv the io_error flag */ - if (f != -1 && remote_version >= 17 && !read_batch) { /* dw-added readbatch */ - extern int module_id; - extern int ignore_errors; - if (lp_ignore_errors(module_id) || ignore_errors) { - read_int(f); - } else { - io_error |= read_int(f); + if (!read_batch) { + /* Recv the io_error flag */ + if (lp_ignore_errors(module_id) || ignore_errors) + read_int(f); + else + io_error |= read_int(f); } } + if (verbose > 3) + output_flist(flist); + if (list_only) { int i; - for (i = 0; i < flist->count; i++) { + for (i = 0; i < flist->count; i++) list_file_entry(flist->files[i]); - } } - if (verbose > 2) rprintf(FINFO, "recv_file_list done\n"); @@ -1086,21 +1297,20 @@ struct file_list *recv_file_list(int f) } -/* - * XXX: This is currently the hottest function while building the file - * list, because building f_name()s every time is expensive. - **/ -int file_compare(struct file_struct **f1, struct file_struct **f2) +int file_compare(struct file_struct **file1, struct file_struct **file2) { - if (!(*f1)->basename && !(*f2)->basename) + struct file_struct *f1 = *file1; + struct file_struct *f2 = *file2; + + if (!f1->basename && !f2->basename) return 0; - if (!(*f1)->basename) + if (!f1->basename) return -1; - if (!(*f2)->basename) + if (!f2->basename) return 1; - if ((*f1)->dirname == (*f2)->dirname) - return u_strcmp((*f1)->basename, (*f2)->basename); - return u_strcmp(f_name(*f1), f_name(*f2)); + if (f1->dirname == f2->dirname) + return u_strcmp(f1->basename, f2->basename); + return f_name_cmp(f1, f2); } @@ -1108,20 +1318,20 @@ int flist_find(struct file_list *flist, struct file_struct *f) { int low = 0, high = flist->count - 1; - if (flist->count <= 0) + while (high >= 0 && !flist->files[high]->basename) high--; + + if (high < 0) return -1; while (low != high) { int mid = (low + high) / 2; - int ret = - file_compare(&flist->files[flist_up(flist, mid)], &f); + int ret = file_compare(&flist->files[flist_up(flist, mid)],&f); if (ret == 0) return flist_up(flist, mid); - if (ret > 0) { + if (ret > 0) high = mid; - } else { + else low = mid + 1; - } } if (file_compare(&flist->files[flist_up(flist, low)], &f) == 0) @@ -1129,44 +1339,43 @@ int flist_find(struct file_list *flist, struct file_struct *f) return -1; } - /* - * free up one file + * Free up any resources a file_struct has allocated + * and clear the file. */ -void free_file(struct file_struct *file) +void clear_file(int i, struct file_list *flist) { - if (!file) - return; - if (file->basename) - free(file->basename); - if (file->link) - free(file->link); - if (file->sum) - free(file->sum); - *file = null_file; + if (flist->hlink_pool && flist->files[i]->link_u.idev) + pool_free(flist->hlink_pool, 0, flist->files[i]->link_u.idev); + memset(flist->files[i], 0, file_struct_len); } /* * allocate a new file list */ -struct file_list *flist_new() +struct file_list *flist_new(int with_hlink, char *msg) { struct file_list *flist; - flist = (struct file_list *) malloc(sizeof(flist[0])); + flist = new(struct file_list); if (!flist) - out_of_memory("send_file_list"); + out_of_memory(msg); - flist->count = 0; - flist->malloced = 0; - flist->files = NULL; + memset(flist, 0, sizeof (struct file_list)); -#if ARENA_SIZE > 0 - flist->string_area = string_area_new(0); -#else - flist->string_area = NULL; + if (!(flist->file_pool = pool_create(FILE_EXTENT, 0, + out_of_memory, POOL_INTERN))) + out_of_memory(msg); + +#if SUPPORT_HARD_LINKS + if (with_hlink && preserve_hard_links) { + if (!(flist->hlink_pool = pool_create(HLINK_EXTENT, + sizeof (struct idev), out_of_memory, POOL_INTERN))) + out_of_memory(msg); + } #endif + return flist; } @@ -1175,64 +1384,53 @@ struct file_list *flist_new() */ void flist_free(struct file_list *flist) { - int i; - for (i = 1; i < flist->count; i++) { - if (!flist->string_area) - free_file(flist->files[i]); - free(flist->files[i]); - } - /* FIXME: I don't think we generally need to blank the flist - * since it's about to be freed. This will just cause more - * memory traffic. If you want a freed-memory debugger, you - * know where to get it. */ - memset((char *) flist->files, 0, - sizeof(flist->files[0]) * flist->count); + pool_destroy(flist->file_pool); + pool_destroy(flist->hlink_pool); free(flist->files); - if (flist->string_area) - string_area_free(flist->string_area); - memset((char *) flist, 0, sizeof(*flist)); 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 + * duplicate names can cause corruption because of the pipelining */ -static void clean_flist(struct file_list *flist, int strip_root) +static void clean_flist(struct file_list *flist, int strip_root, int no_dups) { - int i; + int i, prev_i = 0; if (!flist || flist->count == 0) return; qsort(flist->files, flist->count, - sizeof(flist->files[0]), (int (*)()) file_compare); - - for (i = 1; i < flist->count; i++) { - if (flist->files[i]->basename && - flist->files[i - 1]->basename && - strcmp(f_name(flist->files[i]), - f_name(flist->files[i - 1])) == 0) { - if (verbose > 1 && !am_server) + sizeof flist->files[0], (int (*)()) file_compare); + + for (i = no_dups? 0 : flist->count; i < flist->count; i++) { + if (flist->files[i]->basename) { + prev_i = i; + break; + } + } + while (++i < flist->count) { + if (!flist->files[i]->basename) + continue; + if (f_name_cmp(flist->files[i], flist->files[prev_i]) == 0) { + if (verbose > 1 && !am_server) { rprintf(FINFO, "removing duplicate name %s from file list %d\n", - f_name(flist->files[i - 1]), - i - 1); - /* it's not great that the flist knows the semantics of the - * file memory usage, but i'd rather not add a flag byte - * to that struct. XXX can i use a bit in the flags field? */ - if (flist->string_area) - flist->files[i][0] = null_file; - else - free_file(flist->files[i]); - } + f_name(flist->files[i]), i); + } + /* Make sure that if we unduplicate '.', that we don't + * lose track of a user-specified starting point (or + * else deletions will mysteriously fail with -R). */ + if (flist->files[i]->flags & FLAG_TOP_DIR) + flist->files[prev_i]->flags |= FLAG_TOP_DIR; + + clear_file(i, flist); + } else + prev_i = i; } - /* FIXME: There is a bug here when filenames are repeated more - * than once, because we don't handle freed files when doing - * the comparison. */ - if (strip_root) { /* we need to strip off the root directory in the case of relative paths, but this must be done _after_ @@ -1251,49 +1449,130 @@ static void clean_flist(struct file_list *flist, int strip_root) } } } +} - - if (verbose <= 3) - return; +static void output_flist(struct file_list *flist) +{ + char uidbuf[16], gidbuf[16]; + struct file_struct *file; + int i; for (i = 0; i < flist->count; i++) { - rprintf(FINFO, "[%d] i=%d %s %s mode=0%o len=%.0f\n", - (int) getpid(), i, - NS(flist->files[i]->dirname), - NS(flist->files[i]->basename), - (int) flist->files[i]->mode, - (double) flist->files[i]->length); + file = flist->files[i]; + if (am_root && preserve_uid) + sprintf(uidbuf, " uid=%ld", (long)file->uid); + else + *uidbuf = '\0'; + if (preserve_gid && file->gid != GID_NONE) + sprintf(gidbuf, " gid=%ld", (long)file->gid); + else + *gidbuf = '\0'; + rprintf(FINFO, "[%s] i=%d %s %s %s mode=0%o len=%.0f%s%s\n", + who_am_i(), i, NS(file->basedir), NS(file->dirname), + NS(file->basename), (int) file->mode, + (double) file->length, uidbuf, gidbuf); } } -/* - * return the full filename of a flist entry - * - * This function is too expensive at the moment, because it copies - * strings when often we only want to compare them. In any case, - * using strlcat is silly because it will walk the string repeatedly. +enum fnc_state { fnc_DIR, fnc_SLASH, fnc_BASE }; + +/* Compare the names of two file_struct entities, just like strcmp() + * would do if it were operating on the joined strings. We assume + * that there are no 0-length strings. */ -char *f_name(struct file_struct *f) +int f_name_cmp(struct file_struct *f1, struct file_struct *f2) { - static char names[10][MAXPATHLEN]; - static int n; - char *p = names[n]; + int dif; + const uchar *c1, *c2; + enum fnc_state state1, state2; + + if (!f1 || !f1->basename) { + if (!f2 || !f2->basename) + return 0; + return -1; + } + if (!f2 || !f2->basename) + return 1; + + if (!(c1 = (uchar*)f1->dirname)) { + state1 = fnc_BASE; + c1 = (uchar*)f1->basename; + } else + state1 = fnc_DIR; + if (!(c2 = (uchar*)f2->dirname)) { + state2 = fnc_BASE; + c2 = (uchar*)f2->basename; + } else + state2 = fnc_DIR; + + while (1) { + if ((dif = (int)*c1 - (int)*c2) != 0) + break; + if (!*++c1) { + switch (state1) { + case fnc_DIR: + state1 = fnc_SLASH; + c1 = (uchar*)"/"; + break; + case fnc_SLASH: + state1 = fnc_BASE; + c1 = (uchar*)f1->basename; + break; + case fnc_BASE: + break; + } + } + if (!*++c2) { + switch (state2) { + case fnc_DIR: + state2 = fnc_SLASH; + c2 = (uchar*)"/"; + break; + case fnc_SLASH: + state2 = fnc_BASE; + c2 = (uchar*)f2->basename; + break; + case fnc_BASE: + if (!*c1) + return 0; + break; + } + } + } + return dif; +} + + +/* Return a copy of the full filename of a flist entry, using the indicated + * buffer. No size-checking is done because we checked the size when creating + * the file_struct entry. + */ +char *f_name_to(struct file_struct *f, char *fbuf) +{ if (!f || !f->basename) return NULL; - n = (n + 1) % 10; - if (f->dirname) { - int off; + int len = strlen(f->dirname); + memcpy(fbuf, f->dirname, len); + fbuf[len] = '/'; + strcpy(fbuf + len + 1, f->basename); + } else + strcpy(fbuf, f->basename); + return fbuf; +} - off = strlcpy(p, f->dirname, MAXPATHLEN); - off += strlcpy(p + off, "/", MAXPATHLEN - off); - off += strlcpy(p + off, f->basename, MAXPATHLEN - off); - } else { - strlcpy(p, f->basename, MAXPATHLEN); - } - return p; +/* Like f_name_to(), but we rotate through 5 static buffers of our own. + */ +char *f_name(struct file_struct *f) +{ + static char names[5][MAXPATHLEN]; + static unsigned int n; + + n = (n + 1) % (sizeof names / sizeof names[0]); + + return f_name_to(f, names[n]); }