/*
Copyright (C) Andrew Tridgell 1996
Copyright (C) Paul Mackerras 1996
+ Copyright (C) 2001, 2002 by Martin Pool <mbp@samba.org>
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
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
-/* generate and receive file lists */
+/** @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.
+ *
+ **/
#include "rsync.h"
extern struct stats stats;
-extern int csum_length;
-
extern int verbose;
extern int am_server;
extern int always_checksum;
extern int preserve_times;
extern int relative_paths;
extern int copy_links;
+extern int copy_unsafe_links;
extern int remote_version;
extern int io_error;
+extern int sanitize_paths;
+
+extern int read_batch;
+extern int write_batch;
+
+static char topsrcname[MAXPATHLEN];
static struct exclude_struct **local_exclude_list;
+static struct file_struct null_file;
+
static void clean_flist(struct file_list *flist, int strip_root);
+
+static int show_build_progress_p(void)
+{
+ extern int do_progress;
+
+ return do_progress && verbose && recurse && !am_server;
+}
+
+/**
+ * True if we're local, etc, and should emit progress emssages.
+ **/
+static void emit_build_progress(const struct file_list *flist)
+{
+ rprintf(FINFO,
+ " %d files...\r",
+ flist->count);
+}
+
+
+static void finish_build_progress(const struct file_list *flist)
+{
+ if (verbose && recurse && !am_server) {
+ /* This overwrites the progress line, if any. */
+ rprintf(FINFO, RSYNC_NAME ": %d files to consider.\n",
+ flist->count);
+ }
+}
+
+
+static struct string_area *string_area_new(int size)
+{
+ 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;
+}
+
+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)
+{
+ 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;
+}
+
+static char *string_area_strdup(struct string_area **ap, const char *src)
+{
+ char* dest = string_area_malloc(ap, strlen(src) + 1);
+ return strcpy(dest, src);
+}
+
+static void list_file_entry(struct file_struct *f)
+{
+ char perms[11];
+
+ if (!f->basename)
+ /* this can happen if duplicate names were removed */
+ return;
+
+ permstring(perms, f->mode);
+
+ 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 {
+ 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)
+{
+#if SUPPORT_LINKS
+ 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) {
+ return -1;
+ }
+ Linkbuf[l] = 0;
+ if (copy_unsafe_links && (topsrcname[0] != '\0') &&
+ unsafe_symlink(Linkbuf, topsrcname)) {
+ return do_stat(Path, Buffer);
+ }
+ }
+ return 0;
+#else
+ return do_stat(Path, Buffer);
+#endif
+}
+
int link_stat(const char *Path, STRUCT_STAT *Buffer)
{
#if SUPPORT_LINKS
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
*/
-static int match_file_name(char *fname,STRUCT_STAT *st)
+static int check_exclude_file(int f,char *fname,STRUCT_STAT *st)
{
- if (check_exclude(fname,local_exclude_list,st)) {
- if (verbose > 2)
- rprintf(FINFO,"excluding file %s\n",fname);
- return 0;
- }
- return 1;
+ extern int delete_excluded;
+
+ /* f is set to -1 when calculating deletion file list */
+ if ((f == -1) && delete_excluded) {
+ return 0;
+ }
+ if (check_exclude(fname,local_exclude_list,st)) {
+ return 1;
+ }
+ return 0;
}
/* used by the one_file_system code */
}
+static int to_wire_mode(mode_t mode)
+{
+ if (S_ISLNK(mode) && (_S_IFLNK != 0120000)) {
+ return (mode & ~(_S_IFMT)) | 0120000;
+ }
+ return (int)mode;
+}
+
+static mode_t from_wire_mode(int mode)
+{
+ if ((mode & (_S_IFMT)) == 0120000 && (_S_IFLNK != 0120000)) {
+ return (mode & ~(_S_IFMT)) | _S_IFLNK;
+ }
+ return (mode_t)mode;
+}
+
+
static void send_directory(int f,struct file_list *flist,char *dir);
static char *flist_dir;
if (!(flags & SAME_TIME))
write_int(f,(int)file->modtime);
if (!(flags & SAME_MODE))
- write_int(f,(int)file->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 SUPPORT_HARD_LINKS
if (preserve_hard_links && S_ISREG(file->mode)) {
- write_int(f,(int)file->dev);
- write_int(f,(int)file->inode);
+ if (remote_version < 26) {
+ /* 32-bit dev_t and ino_t */
+ write_int(f,(int)file->dev);
+ write_int(f,(int)file->inode);
+ } else {
+ /* 64-bit dev_t and ino_t */
+ write_longint(f, file->dev);
+ write_longint(f, file->inode);
+ }
}
#endif
if (always_checksum) {
- write_buf(f,file->sum,csum_length);
+ if (remote_version < 21) {
+ write_buf(f,file->sum,2);
+ } else {
+ write_buf(f,file->sum,MD4_SUM_LENGTH);
+ }
}
last_mode = file->mode;
last_gid = file->gid;
last_time = file->modtime;
- strlcpy(lastname,fname,MAXPATHLEN-1);
+ strlcpy(lastname,fname,MAXPATHLEN);
lastname[MAXPATHLEN-1] = 0;
}
static gid_t last_gid;
static char lastname[MAXPATHLEN];
char thisname[MAXPATHLEN];
- int l1=0,l2=0;
+ unsigned int l1=0,l2=0;
char *p;
struct file_struct *file;
memset((char *)file, 0, sizeof(*file));
(*fptr) = file;
- if (l2 >= MAXPATHLEN-l1) overflow("receive_file_entry");
+ if (l2 >= MAXPATHLEN-l1) {
+ rprintf(FERROR,"overflow: flags=0x%x l1=%d l2=%d lastname=%s\n",
+ flags, l1, l2, lastname);
+ overflow("receive_file_entry");
+ }
- strlcpy(thisname,lastname,l1);
+ strlcpy(thisname,lastname,l1+1);
read_sbuf(f,&thisname[l1],l2);
thisname[l1+l2] = 0;
- strlcpy(lastname,thisname,MAXPATHLEN-1);
+ strlcpy(lastname,thisname,MAXPATHLEN);
lastname[MAXPATHLEN-1] = 0;
clean_fname(thisname);
+ if (sanitize_paths) {
+ sanitize_path(thisname, NULL);
+ }
+
if ((p = strrchr(thisname,'/'))) {
static char *lastdir;
*p = 0;
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 : (mode_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)
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);
+ }
}
#if SUPPORT_HARD_LINKS
if (preserve_hard_links && S_ISREG(file->mode)) {
- file->dev = read_int(f);
- file->inode = read_int(f);
+ if (remote_version < 26) {
+ file->dev = read_int(f);
+ file->inode = read_int(f);
+ } else {
+ file->dev = read_longint(f);
+ file->inode = read_longint(f);
+ }
}
#endif
if (always_checksum) {
file->sum = (char *)malloc(MD4_SUM_LENGTH);
if (!file->sum) out_of_memory("md4 sum");
- read_buf(f,file->sum,csum_length);
+ if (remote_version < 21) {
+ read_buf(f,file->sum,2);
+ } else {
+ read_buf(f,file->sum,MD4_SUM_LENGTH);
+ }
}
last_mode = file->mode;
return (st2.st_dev != filesystem_dev);
}
-static struct file_struct *make_file(char *fname)
+#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)
{
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-1);
+ strlcpy(cleaned_name, fname, MAXPATHLEN);
cleaned_name[MAXPATHLEN-1] = 0;
clean_fname(cleaned_name);
+ if (sanitize_paths) {
+ sanitize_path(cleaned_name, NULL);
+ }
fname = cleaned_name;
memset(sum,0,SUM_LENGTH);
- if (link_stat(fname,&st) != 0) {
+ if (readlink_stat(fname,&st,linkbuf) != 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)) {
+ /* file is excluded anyway, ignore silently */
+ return NULL;
+ }
+ }
io_error = 1;
- rprintf(FERROR,"%s: %s\n",
- fname,strerror(errno));
+ rprintf(FERROR,"readlink %s: %s\n",
+ fname,strerror(save_errno));
return NULL;
}
+ /* we use noexcludes from backup.c */
+ if (noexcludes) goto skip_excludes;
+
if (S_ISDIR(st.st_mode) && !recurse) {
rprintf(FINFO,"skipping directory %s\n",fname);
return NULL;
return NULL;
}
- if (!match_file_name(fname,&st))
+ if (check_exclude_file(f,fname,&st))
return NULL;
-
+
+
+ if (lp_ignore_nonreadable(module_id) && access(fname, R_OK) != 0)
+ return NULL;
+
+ skip_excludes:
+
if (verbose > 2)
- rprintf(FINFO,"make_file(%s)\n",fname);
+ rprintf(FINFO,"make_file(%d,%s)\n",f,fname);
file = (struct file_struct *)malloc(sizeof(*file));
if (!file) out_of_memory("make_file");
file->dirname = strdup(fname);
lastdir = file->dirname;
}
- file->basename = strdup(p+1);
+ file->basename = STRDUP(ap, p+1);
*p = '/';
} else {
file->dirname = NULL;
- file->basename = strdup(fname);
+ file->basename = STRDUP(ap, fname);
}
file->modtime = st.st_mtime;
#if SUPPORT_LINKS
if (S_ISLNK(st.st_mode)) {
- int l;
- char lnk[MAXPATHLEN];
- if ((l=readlink(fname,lnk,MAXPATHLEN-1)) == -1) {
- io_error=1;
- rprintf(FERROR,"readlink %s : %s\n",
- fname,strerror(errno));
- return NULL;
- }
- lnk[l] = 0;
- file->link = strdup(lnk);
+ file->link = STRDUP(ap, linkbuf);
}
#endif
if (always_checksum) {
- file->sum = (char *)malloc(MD4_SUM_LENGTH);
+ 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
{
struct file_struct *file;
- file = make_file(fname);
+ file = make_file(f,fname, &flist->string_area, 0);
+
+ if (!file) return;
- if (!file) return;
+ if (show_build_progress_p() & !(flist->count % 100))
+ emit_build_progress(flist);
if (flist->count >= flist->malloced) {
if (flist->malloced < 1000)
out_of_memory("send_file_name");
}
+ if (write_batch) /* dw */
+ file->flags = FLAG_DELETE;
+
if (strcmp(file->basename,"")) {
flist->files[flist->count++] = file;
send_file_entry(file,f,base_flags);
return;
}
- strlcpy(fname,dir,MAXPATHLEN-1);
+ strlcpy(fname,dir,MAXPATHLEN);
l = strlen(fname);
if (fname[l-1] != '/') {
if (l == MAXPATHLEN-1) {
closedir(d);
return;
}
- strlcat(fname,"/", MAXPATHLEN-1);
+ strlcat(fname,"/", MAXPATHLEN);
l++;
}
p = fname + strlen(fname);
if (strcmp(dname,".")==0 ||
strcmp(dname,"..")==0)
continue;
- strlcpy(p,dname,MAXPATHLEN-(l+1));
+ strlcpy(p,dname,MAXPATHLEN-l);
send_file_name(f,flist,fname,recurse,0);
}
}
-
+/*
+ *
+ * I *think* f==-1 means that the list should just be built in memory
+ * and not transmitted. But who can tell? -- mbp
+ */
struct file_list *send_file_list(int f,int argc,char *argv[])
{
int i,l;
int64 start_write;
if (verbose && recurse && !am_server && f != -1) {
- rprintf(FINFO,"building file list ... ");
+ rprintf(FINFO, RSYNC_NAME ": building file list...\n");
+ if (verbose > 1)
+ rprintf(FINFO, "\n");
rflush(FINFO);
}
start_write = stats.total_written;
- flist = (struct file_list *)malloc(sizeof(flist[0]));
- if (!flist) out_of_memory("send_file_list");
-
- flist->count=0;
- flist->malloced = 1000;
- flist->files = (struct file_struct **)malloc(sizeof(flist->files[0])*
- flist->malloced);
- if (!flist->files) out_of_memory("send_file_list");
+ flist = flist_new();
if (f != -1) {
io_start_buffering(f);
}
for (i=0;i<argc;i++) {
- char fname2[MAXPATHLEN];
- char *fname = fname2;
+ char *fname = topsrcname;
- strlcpy(fname,argv[i],MAXPATHLEN-1);
+ strlcpy(fname,argv[i],MAXPATHLEN);
l = strlen(fname);
if (l != 1 && fname[l-1] == '/') {
- strlcat(fname,".",MAXPATHLEN-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 (link_stat(fname,&st) != 0) {
- io_error=1;
- rprintf(FERROR,"%s : %s\n",fname,strerror(errno));
+ if (f != -1) {
+ io_error=1;
+ rprintf(FERROR,"link_stat %s : %s\n",fname,strerror(errno));
+ }
continue;
}
thus getting their permissions right */
*p = 0;
if (strcmp(lastpath,fname)) {
- strlcpy(lastpath, fname, sizeof(lastpath)-1);
+ 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 = 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 = '/';
}
} else {
if (one_file_system)
set_filesystem(fname);
- if (!recurse || !send_included_file_names(f,flist))
- send_file_name(f,flist,fname,recurse,FLAG_DELETE);
+ send_file_name(f,flist,fname,recurse,FLAG_DELETE);
if (olddir != NULL) {
flist_dir = NULL;
if (pop_dir(olddir) != 0) {
rprintf(FERROR,"pop_dir %s : %s\n",
dir,strerror(errno));
- exit_cleanup(1);
+ exit_cleanup(RERR_FILESELECT);
}
}
}
+ topsrcname[0] = '\0';
+
if (f != -1) {
send_file_entry(NULL,f,0);
}
- if (verbose && recurse && !am_server && f != -1)
- rprintf(FINFO,"done\n");
+ finish_build_progress(flist);
clean_flist(flist, 0);
/* if protocol version is >= 17 then send the io_error flag */
if (f != -1 && remote_version >= 17) {
- write_int(f, io_error);
+ extern int module_id;
+ write_int(f, lp_ignore_errors(module_id)? 0 : io_error);
}
if (f != -1) {
io_end_buffering(f);
stats.flist_size = stats.total_written - start_write;
stats.num_files = flist->count;
+ if (write_batch) /* dw */
+ write_batch_flist_info(flist->count, flist->files);
}
if (verbose > 2)
struct file_list *flist;
unsigned char flags;
int64 start_read;
+ extern int list_only;
if (verbose && recurse && !am_server) {
rprintf(FINFO,"receiving file list ... ");
}
/* if protocol version is >= 17 then recv the io_error flag */
- if (f != -1 && remote_version >= 17) {
- io_error |= read_int(f);
+ 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 (list_only) {
+ int i;
+ for (i=0;i<flist->count;i++) {
+ list_file_entry(flist->files[i]);
+ }
}
+
if (verbose > 2)
rprintf(FINFO,"recv_file_list done\n");
}
+/*
+ * 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)
{
if (!(*f1)->basename && !(*f2)->basename) return 0;
/*
* free up one file
*/
-static void free_file(struct file_struct *file)
+void free_file(struct file_struct *file)
{
if (!file) return;
if (file->basename) free(file->basename);
if (file->link) free(file->link);
if (file->sum) free(file->sum);
- memset((char *)file, 0, sizeof(*file));
+ *file = null_file;
}
+/*
+ * allocate a new file list
+ */
+struct file_list *flist_new()
+{
+ struct file_list *flist;
+
+ flist = (struct file_list *)malloc(sizeof(flist[0]));
+ if (!flist) out_of_memory("send_file_list");
+
+ flist->count=0;
+ flist->malloced = 1000;
+ flist->files = (struct file_struct **)malloc(sizeof(flist->files[0])*
+ flist->malloced);
+ if (!flist->files) out_of_memory("send_file_list");
+#if ARENA_SIZE > 0
+ flist->string_area = string_area_new(0);
+#else
+ flist->string_area = NULL;
+#endif
+ return flist;
+}
/*
* free up all elements in a flist
*/
{
int i;
for (i=1;i<flist->count;i++) {
- free_file(flist->files[i]);
+ if (!flist->string_area)
+ free_file(flist->files[i]);
free(flist->files[i]);
}
memset((char *)flist->files, 0, sizeof(flist->files[0])*flist->count);
free(flist->files);
+ if (flist->string_area)
+ string_area_free(flist->string_area);
memset((char *)flist, 0, sizeof(*flist));
free(flist);
}
if (verbose > 1 && !am_server)
rprintf(FINFO,"removing duplicate name %s from file list %d\n",
f_name(flist->files[i-1]),i-1);
- free_file(flist->files[i]);
+ /* 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]);
}
}
+ /* 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_
if (verbose <= 3) return;
for (i=0;i<flist->count;i++) {
- rprintf(FINFO,"[%d] i=%d %s %s mode=0%o len=%d\n",
- getpid(), i,
- flist->files[i]->dirname,
- flist->files[i]->basename,
- flist->files[i]->mode,
- flist->files[i]->length);
+ 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);
}
}
/*
* 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.
*/
char *f_name(struct file_struct *f)
{
n = (n+1)%10;
if (f->dirname) {
- slprintf(p, MAXPATHLEN-1, "%s/%s", f->dirname, f->basename);
+ int off;
+
+ 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-1);
+ strlcpy(p, f->basename, MAXPATHLEN);
}
return p;