X-Git-Url: https://mattmccutchen.net/rsync/rsync.git/blobdiff_plain/c579310a009dc21a0f2b378c711980336f79add6..58b1999e086721ff77d8f3d5e0d29e1e58c70a94:/util.c diff --git a/util.c b/util.c index c1e04710..a2e8660b 100644 --- a/util.c +++ b/util.c @@ -1,42 +1,51 @@ /* -*- c-file-style: "linux" -*- - - Copyright (C) 1996-2000 by Andrew Tridgell - 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. -*/ - -/* - Utilities used in rsync - - tridge, June 1996 - */ + * + * Copyright (C) 1996-2000 by Andrew Tridgell + * 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. + */ + +/** + * @file + * + * Utilities used in rsync + **/ + #include "rsync.h" extern int verbose; +extern int dry_run; +extern int module_id; +extern int modify_window; +extern char *partial_dir; +extern struct exclude_list_struct server_exclude_list; + +int sanitize_paths = 0; -/**************************************************************************** -Set a fd into nonblocking mode -****************************************************************************/ + +/** + * Set a fd into nonblocking mode + **/ void set_nonblocking(int fd) { int val; - if((val = fcntl(fd, F_GETFL, 0)) == -1) + if ((val = fcntl(fd, F_GETFL, 0)) == -1) return; if (!(val & NONBLOCK_FLAG)) { val |= NONBLOCK_FLAG; @@ -44,14 +53,14 @@ void set_nonblocking(int fd) } } -/**************************************************************************** -Set a fd into blocking mode -****************************************************************************/ +/** + * Set a fd into blocking mode + **/ void set_blocking(int fd) { int val; - if((val = fcntl(fd, F_GETFL, 0)) == -1) + if ((val = fcntl(fd, F_GETFL, 0)) == -1) return; if (val & NONBLOCK_FLAG) { val &= ~NONBLOCK_FLAG; @@ -60,10 +69,11 @@ void set_blocking(int fd) } -/* create a file descriptor pair - like pipe() but use socketpair if - possible (because of blocking issues on pipes) - - always set non-blocking +/** + * Create a file descriptor pair - like pipe() but use socketpair if + * possible (because of blocking issues on pipes). + * + * Always set non-blocking. */ int fd_pair(int fd[2]) { @@ -79,12 +89,12 @@ int fd_pair(int fd[2]) set_nonblocking(fd[0]); set_nonblocking(fd[1]); } - + return ret; } -static void print_child_argv(char **cmd) +void print_child_argv(char **cmd) { rprintf(FINFO, "opening connection using "); for (; *cmd; cmd++) { @@ -104,157 +114,34 @@ static void print_child_argv(char **cmd) } -/* this is derived from CVS code - - note that in the child STDIN is set to blocking and STDOUT - is set to non-blocking. This is necessary as rsh relies on stdin being blocking - and ssh relies on stdout being non-blocking - - if blocking_io is set then use blocking io on both fds. That can be - used to cope with badly broken rsh implementations like the one on - solaris. - */ -pid_t piped_child(char **command, int *f_in, int *f_out) -{ - pid_t pid; - int to_child_pipe[2]; - int from_child_pipe[2]; - extern int blocking_io; - - if (verbose >= 2) { - print_child_argv(command); - } - - if (fd_pair(to_child_pipe) < 0 || fd_pair(from_child_pipe) < 0) { - rprintf(FERROR, "pipe: %s\n", strerror(errno)); - exit_cleanup(RERR_IPC); - } - - - pid = do_fork(); - if (pid == -1) { - rprintf(FERROR, "fork: %s\n", strerror(errno)); - exit_cleanup(RERR_IPC); - } - - if (pid == 0) { - extern int orig_umask; - if (dup2(to_child_pipe[0], STDIN_FILENO) < 0 || - close(to_child_pipe[1]) < 0 || - close(from_child_pipe[0]) < 0 || - dup2(from_child_pipe[1], STDOUT_FILENO) < 0) { - rprintf(FERROR, "Failed to dup/close : %s\n", - strerror(errno)); - exit_cleanup(RERR_IPC); - } - if (to_child_pipe[0] != STDIN_FILENO) - close(to_child_pipe[0]); - if (from_child_pipe[1] != STDOUT_FILENO) - close(from_child_pipe[1]); - umask(orig_umask); - set_blocking(STDIN_FILENO); - if (blocking_io) { - set_blocking(STDOUT_FILENO); - } - execvp(command[0], command); - rprintf(FERROR, "Failed to exec %s : %s\n", - command[0], strerror(errno)); - exit_cleanup(RERR_IPC); - } - - if (close(from_child_pipe[1]) < 0 || close(to_child_pipe[0]) < 0) { - rprintf(FERROR, "Failed to close : %s\n", strerror(errno)); - exit_cleanup(RERR_IPC); - } - - *f_in = from_child_pipe[0]; - *f_out = to_child_pipe[1]; - - return pid; -} - -pid_t local_child(int argc, char **argv,int *f_in,int *f_out) -{ - pid_t pid; - int to_child_pipe[2]; - int from_child_pipe[2]; - extern int read_batch; /* dw */ - - if (fd_pair(to_child_pipe) < 0 || - fd_pair(from_child_pipe) < 0) { - rprintf(FERROR,"pipe: %s\n",strerror(errno)); - exit_cleanup(RERR_IPC); - } - - - pid = do_fork(); - if (pid == -1) { - rprintf(FERROR,"fork: %s\n",strerror(errno)); - exit_cleanup(RERR_IPC); - } - - if (pid == 0) { - extern int am_sender; - extern int am_server; - - am_sender = read_batch ? 0 : !am_sender; - am_server = 1; - - if (dup2(to_child_pipe[0], STDIN_FILENO) < 0 || - close(to_child_pipe[1]) < 0 || - close(from_child_pipe[0]) < 0 || - dup2(from_child_pipe[1], STDOUT_FILENO) < 0) { - rprintf(FERROR,"Failed to dup/close : %s\n",strerror(errno)); - exit_cleanup(RERR_IPC); - } - if (to_child_pipe[0] != STDIN_FILENO) close(to_child_pipe[0]); - if (from_child_pipe[1] != STDOUT_FILENO) close(from_child_pipe[1]); - start_server(STDIN_FILENO, STDOUT_FILENO, argc, argv); - } - - if (close(from_child_pipe[1]) < 0 || - close(to_child_pipe[0]) < 0) { - rprintf(FERROR,"Failed to close : %s\n",strerror(errno)); - exit_cleanup(RERR_IPC); - } - - *f_in = from_child_pipe[0]; - *f_out = to_child_pipe[1]; - - return pid; -} - - - void out_of_memory(char *str) { - rprintf(FERROR,"ERROR: out of memory in %s\n",str); - exit_cleanup(RERR_MALLOC); + rprintf(FERROR, "ERROR: out of memory in %s\n", str); + exit_cleanup(RERR_MALLOC); } void overflow(char *str) { - rprintf(FERROR,"ERROR: buffer overflow in %s\n",str); - exit_cleanup(RERR_MALLOC); + rprintf(FERROR, "ERROR: buffer overflow in %s\n", str); + exit_cleanup(RERR_MALLOC); } int set_modtime(char *fname, time_t modtime) { - extern int dry_run; if (dry_run) return 0; if (verbose > 2) { rprintf(FINFO, "set modtime of %s to (%ld) %s", - fname, (long) modtime, + fname, (long)modtime, asctime(localtime(&modtime))); } - + { #ifdef HAVE_UTIMBUF - struct utimbuf tbuf; + struct utimbuf tbuf; tbuf.actime = time(NULL); tbuf.modtime = modtime; return utime(fname,&tbuf); @@ -275,23 +162,24 @@ int set_modtime(char *fname, time_t modtime) } -/**************************************************************************** -create any necessary directories in fname. Unfortunately we don't know -what perms to give the directory when this is called so we need to rely -on the umask -****************************************************************************/ -int create_directory_path(char *fname) +/** + Create any necessary directories in fname. Unfortunately we don't know + what perms to give the directory when this is called so we need to rely + on the umask +**/ +int create_directory_path(char *fname, int base_umask) { - extern int orig_umask; char *p; - while (*fname == '/') fname++; - while (strncmp(fname,"./",2)==0) fname += 2; + while (*fname == '/') + fname++; + while (strncmp(fname, "./", 2) == 0) + fname += 2; p = fname; - while ((p=strchr(p,'/'))) { + while ((p = strchr(p,'/')) != NULL) { *p = 0; - do_mkdir(fname,0777 & ~orig_umask); + do_mkdir(fname, 0777 & ~base_umask); *p = '/'; p++; } @@ -299,23 +187,26 @@ int create_directory_path(char *fname) } -/* Write LEN bytes at PTR to descriptor DESC, retrying if interrupted. - Return LEN upon success, write's (negative) error code otherwise. - - derived from GNU C's cccp.c. -*/ +/** + * Write @p len bytes at @p ptr to descriptor @p desc, retrying if + * interrupted. + * + * @retval len upon success + * + * @retval <0 write's (negative) error code + * + * Derived from GNU C's cccp.c. + */ static int full_write(int desc, char *ptr, size_t len) { int total_written; - + total_written = 0; while (len > 0) { - int written = write (desc, ptr, len); + int written = write(desc, ptr, len); if (written < 0) { -#ifdef EINTR if (errno == EINTR) continue; -#endif return written; } total_written += written; @@ -325,31 +216,36 @@ static int full_write(int desc, char *ptr, size_t len) return total_written; } -/* Read LEN bytes at PTR from descriptor DESC, retrying if interrupted. - Return the actual number of bytes read, zero for EOF, or negative - for an error. - derived from GNU C's cccp.c. */ +/** + * Read @p len bytes at @p ptr from descriptor @p desc, retrying if + * interrupted. + * + * @retval >0 the actual number of bytes read + * + * @retval 0 for EOF + * + * @retval <0 for an error. + * + * Derived from GNU C's cccp.c. */ static int safe_read(int desc, char *ptr, size_t len) { int n_chars; - + if (len == 0) return len; - -#ifdef EINTR + do { n_chars = read(desc, ptr, len); } while (n_chars < 0 && errno == EINTR); -#else - n_chars = read(desc, ptr, len); -#endif - + return n_chars; } -/* copy a file - this is used in conjunction with the --temp-dir option */ +/** Copy a file. + * + * This is used in conjunction with the --temp-dir option */ int copy_file(char *source, char *dest, mode_t mode) { int ifd; @@ -359,59 +255,66 @@ int copy_file(char *source, char *dest, mode_t mode) ifd = do_open(source, O_RDONLY, 0); if (ifd == -1) { - rprintf(FERROR,"open %s: %s\n", - source,strerror(errno)); + rsyserr(FERROR, errno, "open %s", full_fname(source)); return -1; } if (robust_unlink(dest) && errno != ENOENT) { - rprintf(FERROR,"unlink %s: %s\n", - dest,strerror(errno)); + rsyserr(FERROR, errno, "unlink %s", full_fname(dest)); return -1; } ofd = do_open(dest, O_WRONLY | O_CREAT | O_TRUNC | O_EXCL, mode); if (ofd == -1) { - rprintf(FERROR,"open %s: %s\n", - dest,strerror(errno)); + rsyserr(FERROR, errno, "open %s", full_fname(dest)); close(ifd); return -1; } - while ((len = safe_read(ifd, buf, sizeof(buf))) > 0) { + while ((len = safe_read(ifd, buf, sizeof buf)) > 0) { if (full_write(ofd, buf, len) < 0) { - rprintf(FERROR,"write %s: %s\n", - dest,strerror(errno)); + rsyserr(FERROR, errno, "write %s", full_fname(dest)); close(ifd); close(ofd); return -1; } } - close(ifd); - close(ofd); - if (len < 0) { - rprintf(FERROR,"read %s: %s\n", - source,strerror(errno)); + rsyserr(FERROR, errno, "read %s", full_fname(source)); + close(ifd); + close(ofd); + return -1; + } + + if (close(ifd) < 0) { + rsyserr(FINFO, errno, "close failed on %s", + full_fname(source)); + } + + if (close(ofd) < 0) { + rsyserr(FERROR, errno, "close failed on %s", + full_fname(dest)); return -1; } return 0; } -/* - Robust unlink: some OS'es (HPUX) refuse to unlink busy files, so - rename to /.rsyncNNN instead. Note that successive rsync runs - will shuffle the filenames around a bit as long as the file is still - busy; this is because this function does not know if the unlink call - is due to a new file coming in, or --delete trying to remove old - .rsyncNNN files, hence it renames it each time. -*/ /* MAX_RENAMES should be 10**MAX_RENAMES_DIGITS */ #define MAX_RENAMES_DIGITS 3 #define MAX_RENAMES 1000 +/** + * Robust unlink: some OS'es (HPUX) refuse to unlink busy files, so + * rename to /.rsyncNNN instead. + * + * Note that successive rsync runs will shuffle the filenames around a + * bit as long as the file is still busy; this is because this function + * does not know if the unlink call is due to a new file coming in, or + * --delete trying to remove old .rsyncNNN files, hence it renames it + * each time. + **/ int robust_unlink(char *fname) { #ifndef ETXTBSY @@ -422,17 +325,15 @@ int robust_unlink(char *fname) char path[MAXPATHLEN]; rc = do_unlink(fname); - if ((rc == 0) || (errno != ETXTBSY)) + if (rc == 0 || errno != ETXTBSY) return rc; - strlcpy(path, fname, MAXPATHLEN); + if ((pos = strlcpy(path, fname, MAXPATHLEN)) >= MAXPATHLEN) + pos = MAXPATHLEN - 1; - pos = strlen(path); - while((path[--pos] != '/') && (pos >= 0)) - ; - ++pos; - strlcpy(&path[pos], ".rsync", MAXPATHLEN-pos); - pos += sizeof(".rsync")-1; + while (pos > 0 && path[pos-1] != '/') + pos--; + pos += strlcpy(path+pos, ".rsync", MAXPATHLEN-pos); if (pos > (MAXPATHLEN-MAX_RENAMES_DIGITS-1)) { errno = ETXTBSY; @@ -445,11 +346,12 @@ int robust_unlink(char *fname) sprintf(&path[pos], "%03d", counter); if (++counter >= MAX_RENAMES) counter = 1; - } while (((rc = access(path, 0)) == 0) && (counter != start)); + } while ((rc = access(path, 0)) == 0 && counter != start); - if (verbose > 0) + if (verbose > 0) { rprintf(FINFO,"renaming %s to %s because of text busy\n", - fname, path); + fname, path); + } /* maybe we should return rename()'s exit status? Nah. */ if (do_rename(fname, path) != 0) { @@ -460,50 +362,86 @@ int robust_unlink(char *fname) #endif } -int robust_rename(char *from, char *to) +/* Returns 0 on successful rename, 1 if we successfully copied the file + * across filesystems, -2 if copy_file() failed, and -1 on other errors. */ +int robust_rename(char *from, char *to, int mode) { -#ifndef ETXTBSY - return do_rename(from, to); -#else - int rc = do_rename(from, to); - if ((rc == 0) || (errno != ETXTBSY)) - return rc; - if (robust_unlink(to) != 0) - return -1; - return do_rename(from, to); + int tries = 4; + + while (tries--) { + if (do_rename(from, to) == 0) + return 0; + + switch (errno) { +#ifdef ETXTBSY + case ETXTBSY: + if (robust_unlink(to) != 0) + return -1; + break; #endif + case EXDEV: + if (copy_file(from, to, mode) != 0) + return -2; + do_unlink(from); + return 1; + default: + return -1; + } + } + return -1; } static pid_t all_pids[10]; static int num_pids; -/* fork and record the pid of the child */ +/** Fork and record the pid of the child. **/ pid_t do_fork(void) { pid_t newpid = fork(); - - if (newpid) { + + if (newpid != 0 && newpid != -1) { all_pids[num_pids++] = newpid; } return newpid; } -/* kill all children */ +/** + * Kill all children. + * + * @todo It would be kind of nice to make sure that they are actually + * all our children before we kill them, because their pids may have + * been recycled by some other process. Perhaps when we wait for a + * child, we should remove it from this array. Alternatively we could + * perhaps use process groups, but I think that would not work on + * ancient Unix versions that don't support them. + **/ void kill_all(int sig) { int i; - for (i=0;ipw_uid; @@ -512,11 +450,12 @@ int name_to_uid(char *name, uid_t *uid) return 0; } -/* turn a group name into a gid */ +/** Turn a group name into a gid */ int name_to_gid(char *name, gid_t *gid) { struct group *grp; - if (!name || !*name) return 0; + if (!name || !*name) + return 0; grp = getgrnam(name); if (grp) { *gid = grp->gr_gid; @@ -526,7 +465,7 @@ int name_to_gid(char *name, gid_t *gid) } -/* lock a byte range in a open file */ +/** Lock a byte range in a open file */ int lock_range(int fd, int offset, int len) { struct flock lock; @@ -536,465 +475,629 @@ int lock_range(int fd, int offset, int len) lock.l_start = offset; lock.l_len = len; lock.l_pid = 0; - + return fcntl(fd,F_SETLK,&lock) == 0; } +static int exclude_server_path(char *arg) +{ + char *s; + + if (server_exclude_list.head) { + for (s = arg; (s = strchr(s, '/')) != NULL; ) { + *s = '\0'; + if (check_exclude(&server_exclude_list, arg, 1) < 0) { + /* We must leave arg truncated! */ + return 1; + } + *s++ = '/'; + } + } + return 0; +} -static void glob_expand_one(char *s, char **argv, int *argc, int maxargs) +static void glob_expand_one(char *s, char ***argv_ptr, int *argc_ptr, + int *maxargs_ptr) { + char **argv = *argv_ptr; + int argc = *argc_ptr; + int maxargs = *maxargs_ptr; #if !(defined(HAVE_GLOB) && defined(HAVE_GLOB_H)) - if (!*s) s = "."; - argv[*argc] = strdup(s); - (*argc)++; - return; + if (argc == maxargs) { + maxargs += MAX_ARGS; + if (!(argv = realloc_array(argv, char *, maxargs))) + out_of_memory("glob_expand_one"); + *argv_ptr = argv; + *maxargs_ptr = maxargs; + } + if (!*s) + s = "."; + s = argv[argc++] = strdup(s); + exclude_server_path(s); #else - extern int sanitize_paths; glob_t globbuf; int i; - if (!*s) s = "."; - - argv[*argc] = strdup(s); - if (sanitize_paths) { - sanitize_path(argv[*argc], NULL); - } - - memset(&globbuf, 0, sizeof(globbuf)); - glob(argv[*argc], 0, NULL, &globbuf); - if (globbuf.gl_pathc == 0) { - (*argc)++; - globfree(&globbuf); + if (maxargs <= argc) return; + if (!*s) + s = "."; + + if (sanitize_paths) + s = sanitize_path(NULL, s, "", 0); + else + s = strdup(s); + + memset(&globbuf, 0, sizeof globbuf); + if (!exclude_server_path(s)) + glob(s, 0, NULL, &globbuf); + if (MAX((int)globbuf.gl_pathc, 1) > maxargs - argc) { + maxargs += globbuf.gl_pathc + MAX_ARGS; + if (!(argv = realloc_array(argv, char *, maxargs))) + out_of_memory("glob_expand_one"); + *argv_ptr = argv; + *maxargs_ptr = maxargs; } - for (i=0; i<(maxargs - (*argc)) && i < (int) globbuf.gl_pathc;i++) { - if (i == 0) free(argv[*argc]); - argv[(*argc) + i] = strdup(globbuf.gl_pathv[i]); - if (!argv[(*argc) + i]) out_of_memory("glob_expand"); + if (globbuf.gl_pathc == 0) + argv[argc++] = s; + else { + int j = globbuf.gl_pathc; + free(s); + for (i = 0; i < j; i++) { + if (!(argv[argc++] = strdup(globbuf.gl_pathv[i]))) + out_of_memory("glob_expand_one"); + } } globfree(&globbuf); - (*argc) += i; #endif + *argc_ptr = argc; } -void glob_expand(char *base1, char **argv, int *argc, int maxargs) +/* This routine is only used in daemon mode. */ +void glob_expand(char *base1, char ***argv_ptr, int *argc_ptr, int *maxargs_ptr) { - char *s = argv[*argc]; + char *s = (*argv_ptr)[*argc_ptr]; char *p, *q; char *base = base1; + int base_len = strlen(base); - if (!s || !*s) return; + if (!s || !*s) + return; - if (strncmp(s, base, strlen(base)) == 0) { - s += strlen(base); - } + if (strncmp(s, base, base_len) == 0) + s += base_len; - s = strdup(s); - if (!s) out_of_memory("glob_expand"); + if (!(s = strdup(s))) + out_of_memory("glob_expand"); - if (asprintf(&base," %s/", base1) <= 0) out_of_memory("glob_expand"); + if (asprintf(&base," %s/", base1) <= 0) + out_of_memory("glob_expand"); + base_len++; - q = s; - while ((p = strstr(q,base)) && ((*argc) < maxargs)) { - /* split it at this point */ - *p = 0; - glob_expand_one(q, argv, argc, maxargs); - q = p+strlen(base); + for (q = s; *q; q = p + base_len) { + if ((p = strstr(q, base)) != NULL) + *p = '\0'; /* split it at this point */ + glob_expand_one(q, argv_ptr, argc_ptr, maxargs_ptr); + if (!p) + break; } - if (*q && (*argc < maxargs)) glob_expand_one(q, argv, argc, maxargs); - free(s); free(base); } -/******************************************************************* - convert a string to lower case -********************************************************************/ +/** + * Convert a string to lower case + **/ void strlower(char *s) { while (*s) { - if (isupper(*s)) *s = tolower(*s); + if (isupper(*(unsigned char *)s)) + *s = tolower(*(unsigned char *)s); s++; } } -void *Realloc(void *p, int size) +/* Join strings p1 & p2 into "dest" with a guaranteed '/' between them. (If + * p1 ends with a '/', no extra '/' is inserted.) Returns the length of both + * strings + 1 (if '/' was inserted), regardless of whether the null-terminated + * string fits into destsize. */ +size_t pathjoin(char *dest, size_t destsize, const char *p1, const char *p2) { - if (!p) return (void *)malloc(size); - return (void *)realloc(p, size); + size_t len = strlcpy(dest, p1, destsize); + if (len < destsize - 1) { + if (!len || dest[len-1] != '/') + dest[len++] = '/'; + if (len < destsize - 1) + len += strlcpy(dest + len, p2, destsize - len); + else { + dest[len] = '\0'; + len += strlen(p2); + } + } + else + len += strlen(p2) + 1; /* Assume we'd insert a '/'. */ + return len; } +/* Join any number of strings together, putting them in "dest". The return + * value is the length of all the strings, regardless of whether the null- + * terminated whole fits in destsize. Your list of string pointers must end + * with a NULL to indicate the end of the list. */ +size_t stringjoin(char *dest, size_t destsize, ...) +{ + va_list ap; + size_t len, ret = 0; + const char *src; + + va_start(ap, destsize); + while (1) { + if (!(src = va_arg(ap, const char *))) + break; + len = strlen(src); + ret += len; + if (destsize > 1) { + if (len >= destsize) + len = destsize - 1; + memcpy(dest, src, len); + destsize -= len; + dest += len; + } + } + *dest = '\0'; + va_end(ap); + + return ret; +} -void clean_fname(char *name) +int count_dir_elements(const char *p) { - char *p; - int l; - int modified = 1; + int cnt = 0, new_component = 1; + while (*p) { + if (*p++ == '/') + new_component = 1; + else if (new_component) { + new_component = 0; + cnt++; + } + } + return cnt; +} - if (!name) return; +/* Turns multiple adjacent slashes into a single slash; gets rid of "./" + * elements; collapses ".." elements except for those at the start of the + * string; removes a trailing slash. If the resulting name would be empty, + * change it into a ".". */ +unsigned int clean_fname(char *name) +{ + char *limit = name - 1, *t = name, *f = name; + int anchored; - while (modified) { - modified = 0; + if (!name) + return 0; - if ((p=strstr(name,"/./"))) { - modified = 1; - while (*p) { - p[0] = p[2]; - p++; - } + if ((anchored = *f == '/') != 0) + *t++ = *f++; + while (*f) { + /* discard extra slashes */ + if (*f == '/') { + f++; + continue; } - - if ((p=strstr(name,"//"))) { - modified = 1; - while (*p) { - p[0] = p[1]; - p++; + if (*f == '.') { + /* discard "." dirs (but NOT a trailing '.'!) */ + if (f[1] == '/') { + f += 2; + continue; + } + /* collapse ".." dirs */ + if (f[1] == '.' && (f[2] == '/' || !f[2])) { + char *s = t - 1; + if (s == name && anchored) { + f += 2; + continue; + } + while (s > limit && *--s != '/') {} + if (s != t - 1 && (s < name || *s == '/')) { + t = s + 1; + f += 2; + continue; + } + *t++ = *f++; + *t++ = *f++; + limit = t; } } + while (*f && (*t++ = *f++) != '/') {} + } - if (strncmp(p=name,"./",2) == 0) { - modified = 1; - do { - p[0] = p[2]; - } while (*p++); - } + if (t > name+anchored && t[-1] == '/') + t--; + if (t == name) + *t++ = '.'; + *t = '\0'; - l = strlen(p=name); - if (l > 1 && p[l-1] == '/') { - modified = 1; - p[l-1] = 0; - } - } + return t - name; } -/* - * Make path appear as if a chroot had occurred: - * 1. remove leading "/" (or replace with "." if at end) - * 2. remove leading ".." components (except those allowed by "reldir") - * 3. delete any other "/.." (recursively) - * Can only shrink paths, so sanitizes in place. - * While we're at it, remove double slashes and "." components like - * clean_fname does(), but DON'T remove a trailing slash because that - * is sometimes significant on command line arguments. - * If "reldir" is non-null, it is a sanitized directory that the path will be - * relative to, so allow as many ".." at the beginning of the path as - * there are components in reldir. This is used for symbolic link targets. - * If reldir is non-null and the path began with "/", to be completely like - * a chroot we should add in depth levels of ".." at the beginning of the - * path, but that would blow the assumption that the path doesn't grow and - * it is not likely to end up being a valid symlink anyway, so just do - * the normal removal of the leading "/" instead. - * Contributed by Dave Dykstra +/* Make path appear as if a chroot had occurred. This handles a leading + * "/" (either removing it or expanding it) and any leading or embedded + * ".." components that attempt to escape past the module's top dir. + * + * If dest is NULL, a buffer is allocated to hold the result. It is legal + * to call with the dest and the path (p) pointing to the same buffer, but + * rootdir will be ignored to avoid expansion of the string. + * + * The rootdir string contains a value to use in place of a leading slash. + * Specify NULL to get the default of lp_path(module_id). + * + * If depth is > 0, it is a count of how many '..'s to allow at the start + * of the path. + * + * We call clean_fname() to clean up the path, but we preserve a trailing + * slash because that is sometimes significant on command-line arguments. */ - -void sanitize_path(char *p, char *reldir) +char *sanitize_path(char *dest, const char *p, const char *rootdir, int depth) { char *start, *sanp; - int depth = 0; - int allowdotdot = 0; - - if (reldir) { - depth++; - while (*reldir) { - if (*reldir++ == '/') { - depth++; - } + int rlen = 0; + + if (dest != p) { + int plen = strlen(p); + if (*p == '/') { + if (!rootdir) + rootdir = lp_path(module_id); + rlen = strlen(rootdir); + depth = 0; + p++; + } + if (dest) { + if (rlen + plen + 1 >= MAXPATHLEN) + return NULL; + } else if (!(dest = new_array(char, rlen + plen + 1))) + out_of_memory("sanitize_path"); + if (rlen) { + memcpy(dest, rootdir, rlen); + if (rlen > 1) + dest[rlen++] = '/'; } } - start = p; - sanp = p; - while (*p == '/') { - /* remove leading slashes */ - p++; - } + + start = sanp = dest + rlen; while (*p != '\0') { + /* discard leading or extra slashes */ + if (*p == '/') { + p++; + continue; + } /* this loop iterates once per filename component in p. * both p (and sanp if the original had a slash) should * always be left pointing after a slash */ - if ((*p == '.') && ((*(p+1) == '/') || (*(p+1) == '\0'))) { + if (*p == '.' && (p[1] == '/' || p[1] == '\0')) { /* skip "." component */ - while (*++p == '/') { - /* skip following slashes */ - ; - } + p++; continue; } - allowdotdot = 0; - if ((*p == '.') && (*(p+1) == '.') && - ((*(p+2) == '/') || (*(p+2) == '\0'))) { + if (*p == '.' && p[1] == '.' && (p[2] == '/' || p[2] == '\0')) { /* ".." component followed by slash or end */ - if ((depth > 0) && (sanp == start)) { - /* allow depth levels of .. at the beginning */ - --depth; - allowdotdot = 1; - } else { + if (depth <= 0 || sanp != start) { p += 2; - if (*p == '/') - p++; if (sanp != start) { /* back up sanp one level */ --sanp; /* now pointing at slash */ - while ((sanp > start) && (*(sanp - 1) != '/')) { + while (sanp > start && sanp[-1] != '/') { /* skip back up to slash */ sanp--; } } continue; } - } - while (1) { - /* copy one component through next slash */ - *sanp++ = *p++; - if ((*p == '\0') || (*(p-1) == '/')) { - while (*p == '/') { - /* skip multiple slashes */ - p++; - } - break; - } - } - if (allowdotdot) { + /* allow depth levels of .. at the beginning */ + depth--; /* move the virtual beginning to leave the .. alone */ - start = sanp; + start = sanp + 3; } + /* copy one component through next slash */ + while (*p && (*sanp++ = *p++) != '/') {} } - if ((sanp == start) && !allowdotdot) { + if (sanp == dest) { /* ended up with nothing, so put in "." component */ - /* - * note that the !allowdotdot doesn't prevent this from - * happening in all allowed ".." situations, but I didn't - * think it was worth putting in an extra variable to ensure - * it since an extra "." won't hurt in those situations. - */ *sanp++ = '.'; } *sanp = '\0'; -} + return dest; +} -static char curr_dir[MAXPATHLEN]; +char curr_dir[MAXPATHLEN]; +unsigned int curr_dir_len; -/* like chdir() but can be reversed with pop_dir() if save is set. It - is also much faster as it remembers where we have been */ -char *push_dir(char *dir, int save) +/** + * Like chdir(), but it keeps track of the current directory (in the + * global "curr_dir"), and ensures that the path size doesn't overflow. + * Also cleans the path using the clean_fname() function. + **/ +int push_dir(char *dir) { - char *ret = curr_dir; static int initialised; + unsigned int len; if (!initialised) { initialised = 1; - getcwd(curr_dir, sizeof(curr_dir)-1); + getcwd(curr_dir, sizeof curr_dir - 1); + curr_dir_len = strlen(curr_dir); } - if (!dir) return NULL; /* this call was probably just to initialize */ + if (!dir) /* this call was probably just to initialize */ + return 0; - if (chdir(dir)) return NULL; + len = strlen(dir); + if (len == 1 && *dir == '.') + return 1; - if (save) { - ret = strdup(curr_dir); - } + if ((*dir == '/' ? len : curr_dir_len + 1 + len) >= sizeof curr_dir) + return 0; + + if (chdir(dir)) + return 0; if (*dir == '/') { - strlcpy(curr_dir, dir, sizeof(curr_dir)); + memcpy(curr_dir, dir, len + 1); + curr_dir_len = len; } else { - strlcat(curr_dir,"/", sizeof(curr_dir)); - strlcat(curr_dir,dir, sizeof(curr_dir)); + curr_dir[curr_dir_len++] = '/'; + memcpy(curr_dir + curr_dir_len, dir, len + 1); + curr_dir_len += len; } - clean_fname(curr_dir); + curr_dir_len = clean_fname(curr_dir); - return ret; + return 1; } -/* reverse a push_dir call */ +/** + * Reverse a push_dir() call. You must pass in an absolute path + * that was copied from a prior value of "curr_dir". + **/ int pop_dir(char *dir) { - int ret; + if (chdir(dir)) + return 0; - ret = chdir(dir); - if (ret) { - free(dir); - return ret; - } + curr_dir_len = strlcpy(curr_dir, dir, sizeof curr_dir); + if (curr_dir_len >= sizeof curr_dir) + curr_dir_len = sizeof curr_dir - 1; - strlcpy(curr_dir, dir, sizeof(curr_dir)); + return 1; +} - free(dir); +/** + * Return the filename, turning any newlines into '?'s. This ensures that + * outputting it on a line of its own cannot generate an empty line. This + * function can handle only 2 names at a time! + **/ +const char *safe_fname(const char *fname) +{ + static char fbuf1[MAXPATHLEN], fbuf2[MAXPATHLEN]; + static char *fbuf = fbuf2; + char *nl = strchr(fname, '\n'); - return 0; + if (!nl) + return fname; + + fbuf = fbuf == fbuf1 ? fbuf2 : fbuf1; + strlcpy(fbuf, fname, MAXPATHLEN); + nl = fbuf + (nl - (char *)fname); + do { + *nl = '?'; + } while ((nl = strchr(nl+1, '\n')) != NULL); + + return fbuf; } -/* we need to supply our own strcmp function for file list comparisons - to ensure that signed/unsigned usage is consistent between machines. */ -int u_strcmp(const char *cs1, const char *cs2) +/** + * Return a quoted string with the full pathname of the indicated filename. + * The string " (in MODNAME)" may also be appended. The returned pointer + * remains valid until the next time full_fname() is called. + **/ +char *full_fname(const char *fn) { - const uchar *s1 = (const uchar *)cs1; - const uchar *s2 = (const uchar *)cs2; - - while (*s1 && *s2 && (*s1 == *s2)) { - s1++; s2++; + static char *result = NULL; + char *m1, *m2, *m3; + char *p1, *p2; + + if (result) + free(result); + + fn = safe_fname(fn); + if (*fn == '/') + p1 = p2 = ""; + else { + p1 = curr_dir; + p2 = "/"; } - - return (int)*s1 - (int)*s2; -} + if (module_id >= 0) { + m1 = " (in "; + m2 = lp_name(module_id); + m3 = ")"; + if (*p1) { + if (!lp_use_chroot(module_id)) { + char *p = lp_path(module_id); + if (*p != '/' || p[1]) + p1 += strlen(p); + } + if (!*p1) + p2++; + else + p1++; + } + else + fn++; + } else + m1 = m2 = m3 = ""; -static OFF_T last_ofs; -static struct timeval print_time; -static struct timeval start_time; -static OFF_T start_ofs; + asprintf(&result, "\"%s%s%s\"%s%s%s", p1, p2, fn, m1, m2, m3); -static unsigned long msdiff(struct timeval *t1, struct timeval *t2) -{ - return (t2->tv_sec - t1->tv_sec) * 1000 - + (t2->tv_usec - t1->tv_usec) / 1000; + return result; } +static char partial_fname[MAXPATHLEN]; -/** - * @param ofs Current position in file - * @param size Total size of file - * @param is_last True if this is the last time progress will be - * printed for this file, so we should output a newline. (Not - * necessarily the same as all bytes being received.) - **/ -static void rprint_progress(OFF_T ofs, OFF_T size, struct timeval *now, - int is_last) +char *partial_dir_fname(const char *fname) { - int pct = (ofs == size) ? 100 : (int)((100.0*ofs)/size); - unsigned long diff = msdiff(&start_time, now); - double rate = diff ? (double) (ofs-start_ofs) * 1000.0 / diff / 1024.0 : 0; - const char *units; - /* If we've finished transferring this file, show the time taken; - * otherwise show expected time to complete. That's kind of - * inconsistent, but people can probably cope. Hopefully we'll - * get more consistent and complete progress reporting soon. -- - * mbp */ - double remain = is_last - ? (double) diff / 1000.0 - : rate ? (double) (size-ofs) / rate / 1000.0 : 0.0; - int remain_h, remain_m, remain_s; - - if (rate > 1024*1024) { - rate /= 1024.0 * 1024.0; - units = "GB/s"; - } else if (rate > 1024) { - rate /= 1024.0; - units = "MB/s"; - } else { - units = "kB/s"; - } - - remain_s = (int) remain % 60; - remain_m = (int) (remain / 60.0) % 60; - remain_h = (int) (remain / 3600.0); - - rprintf(FINFO, "%12.0f %3d%% %7.2f%s %4d:%02d:%02d%s", - (double) ofs, pct, rate, units, - remain_h, remain_m, remain_s, - is_last ? "\n" : "\r"); + char *t = partial_fname; + int sz = sizeof partial_fname; + const char *fn; + + if ((fn = strrchr(fname, '/')) != NULL) { + fn++; + if (*partial_dir != '/') { + int len = fn - fname; + strncpy(t, fname, len); /* safe */ + t += len; + sz -= len; + } + } else + fn = fname; + if ((int)pathjoin(t, sz, partial_dir, fn) >= sz) + return NULL; + if (server_exclude_list.head + && check_exclude(&server_exclude_list, partial_fname, 0) < 0) + return NULL; + + return partial_fname; } -void end_progress(OFF_T size) +/* If no --partial-dir option was specified, we don't need to do anything + * (the partial-dir is essentially '.'), so just return success. */ +int handle_partial_dir(const char *fname, int create) { - extern int do_progress, am_server; + char *fn, *dir; - if (do_progress && !am_server) { - struct timeval now; - gettimeofday(&now, NULL); - rprint_progress(size, size, &now, True); - } - last_ofs = 0; - start_ofs = 0; - print_time.tv_sec = print_time.tv_usec = 0; - start_time.tv_sec = start_time.tv_usec = 0; + if (fname != partial_fname) + return 1; + if (!create && *partial_dir == '/') + return 1; + if (!(fn = strrchr(partial_fname, '/'))) + return 1; + + *fn = '\0'; + dir = partial_fname; + if (create) { + STRUCT_STAT st; +#if SUPPORT_LINKS + int statret = do_lstat(dir, &st); +#else + int statret = do_stat(dir, &st); +#endif + if (statret == 0 && !S_ISDIR(st.st_mode)) { + if (do_unlink(dir) < 0) + return 0; + statret = -1; + } + if (statret < 0 && do_mkdir(dir, 0700) < 0) + return 0; + } else + do_rmdir(dir); + *fn = '/'; + + return 1; } -void show_progress(OFF_T ofs, OFF_T size) +/** We need to supply our own strcmp function for file list comparisons + to ensure that signed/unsigned usage is consistent between machines. */ +int u_strcmp(const char *cs1, const char *cs2) { - extern int do_progress, am_server; - struct timeval now; - - gettimeofday(&now, NULL); - - if (!start_time.tv_sec && !start_time.tv_usec) { - start_time.tv_sec = now.tv_sec; - start_time.tv_usec = now.tv_usec; - start_ofs = ofs; - } - - if (do_progress - && !am_server - && ofs > last_ofs + 1000 - && msdiff(&print_time, &now) > 250) { - rprint_progress(ofs, size, &now, False); - last_ofs = ofs; - print_time.tv_sec = now.tv_sec; - print_time.tv_usec = now.tv_usec; + const uchar *s1 = (const uchar *)cs1; + const uchar *s2 = (const uchar *)cs2; + + while (*s1 && *s2 && (*s1 == *s2)) { + s1++; s2++; } + + return (int)*s1 - (int)*s2; } -/* determine if a symlink points outside the current directory tree */ -int unsafe_symlink(char *dest, char *src) + + +/** + * Determine if a symlink points outside the current directory tree. + * This is considered "unsafe" because e.g. when mirroring somebody + * else's machine it might allow them to establish a symlink to + * /etc/passwd, and then read it through a web server. + * + * Null symlinks and absolute symlinks are always unsafe. + * + * Basically here we are concerned with symlinks whose target contains + * "..", because this might cause us to walk back up out of the + * transferred directory. We are not allowed to go back up and + * reenter. + * + * @param dest Target of the symlink in question. + * + * @param src Top source directory currently applicable. Basically this + * is the first parameter to rsync in a simple invocation, but it's + * modified by flist.c in slightly complex ways. + * + * @retval True if unsafe + * @retval False is unsafe + * + * @sa t_unsafe.c + **/ +int unsafe_symlink(const char *dest, const char *src) { - char *tok; + const char *name, *slash; int depth = 0; /* all absolute and null symlinks are unsafe */ - if (!dest || !(*dest) || (*dest == '/')) return 1; - - src = strdup(src); - if (!src) out_of_memory("unsafe_symlink"); + if (!dest || !*dest || *dest == '/') + return 1; /* find out what our safety margin is */ - for (tok=strtok(src,"/"); tok; tok=strtok(NULL,"/")) { - if (strcmp(tok,"..") == 0) { - depth=0; - } else if (strcmp(tok,".") == 0) { + for (name = src; (slash = strchr(name, '/')) != 0; name = slash+1) { + if (strncmp(name, "../", 3) == 0) { + depth = 0; + } else if (strncmp(name, "./", 2) == 0) { /* nothing */ } else { depth++; } } - free(src); - - /* drop by one to account for the filename portion */ - depth--; - - dest = strdup(dest); - if (!dest) out_of_memory("unsafe_symlink"); - - for (tok=strtok(dest,"/"); tok; tok=strtok(NULL,"/")) { - if (strcmp(tok,"..") == 0) { - depth--; - } else if (strcmp(tok,".") == 0) { + if (strcmp(name, "..") == 0) + depth = 0; + + for (name = dest; (slash = strchr(name, '/')) != 0; name = slash+1) { + if (strncmp(name, "../", 3) == 0) { + /* if at any point we go outside the current directory + then stop - it is unsafe */ + if (--depth < 0) + return 1; + } else if (strncmp(name, "./", 2) == 0) { /* nothing */ } else { depth++; } - /* if at any point we go outside the current directory then - stop - it is unsafe */ - if (depth < 0) break; } + if (strcmp(name, "..") == 0) + depth--; - free(dest); return (depth < 0); } -/**************************************************************************** - return the date and time as a string -****************************************************************************/ +/** + * Return the date and time as a string + **/ char *timestring(time_t t) { static char TimeBuf[200]; struct tm *tm = localtime(&t); #ifdef HAVE_STRFTIME - strftime(TimeBuf,sizeof(TimeBuf)-1,"%Y/%m/%d %T",tm); + strftime(TimeBuf, sizeof TimeBuf - 1, "%Y/%m/%d %H:%M:%S", tm); #else - strlcpy(TimeBuf, asctime(tm), sizeof(TimeBuf)); + strlcpy(TimeBuf, asctime(tm), sizeof TimeBuf); #endif if (TimeBuf[strlen(TimeBuf)-1] == '\n') { @@ -1013,21 +1116,20 @@ char *timestring(time_t t) **/ int msleep(int t) { - int tdiff=0; - struct timeval tval,t1,t2; + int tdiff = 0; + struct timeval tval, t1, t2; gettimeofday(&t1, NULL); - gettimeofday(&t2, NULL); - + while (tdiff < t) { tval.tv_sec = (t-tdiff)/1000; tval.tv_usec = 1000*((t-tdiff)%1000); - + errno = 0; select(0,NULL,NULL, NULL, &tval); gettimeofday(&t2, NULL); - tdiff = (t2.tv_sec - t1.tv_sec)*1000 + + tdiff = (t2.tv_sec - t1.tv_sec)*1000 + (t2.tv_usec - t1.tv_usec)/1000; } @@ -1035,21 +1137,26 @@ int msleep(int t) } -/******************************************************************* - Determine if two file modification times are equivalent (either exact - or in the modification timestamp window established by --modify-window) - Returns 0 if the times should be treated as the same, 1 if the - first is later and -1 if the 2nd is later - *******************************************************************/ +/** + * Determine if two file modification times are equivalent (either + * exact or in the modification timestamp window established by + * --modify-window). + * + * @retval 0 if the times should be treated as the same + * + * @retval +1 if the first is later + * + * @retval -1 if the 2nd is later + **/ int cmp_modtime(time_t file1, time_t file2) { - extern int modify_window; - if (file2 > file1) { - if (file2 - file1 <= modify_window) return 0; + if (file2 - file1 <= modify_window) + return 0; return -1; } - if (file1 - file2 <= modify_window) return 0; + if (file1 - file2 <= modify_window) + return 0; return 1; } @@ -1057,18 +1164,18 @@ int cmp_modtime(time_t file1, time_t file2) #ifdef __INSURE__XX #include -/******************************************************************* -This routine is a trick to immediately catch errors when debugging -with insure. A xterm with a gdb is popped up when insure catches -a error. It is Linux specific. -********************************************************************/ +/** + This routine is a trick to immediately catch errors when debugging + with insure. A xterm with a gdb is popped up when insure catches + a error. It is Linux specific. +**/ int _Insure_trap_error(int a1, int a2, int a3, int a4, int a5, int a6) { static int (*fn)(); int ret; char *cmd; - asprintf(&cmd, "/usr/X11R6/bin/xterm -display :0 -T Panic -n Panic -e /bin/sh -c 'cat /tmp/ierrs.*.%d ; gdb /proc/%d/exe %d'", + asprintf(&cmd, "/usr/X11R6/bin/xterm -display :0 -T Panic -n Panic -e /bin/sh -c 'cat /tmp/ierrs.*.%d ; gdb /proc/%d/exe %d'", getpid(), getpid(), getpid()); if (!fn) { @@ -1086,3 +1193,23 @@ int _Insure_trap_error(int a1, int a2, int a3, int a4, int a5, int a6) return ret; } #endif + + +#define MALLOC_MAX 0x40000000 + +void *_new_array(unsigned int size, unsigned long num) +{ + if (num >= MALLOC_MAX/size) + return NULL; + return malloc(size * num); +} + +void *_realloc_array(void *ptr, unsigned int size, unsigned long num) +{ + if (num >= MALLOC_MAX/size) + return NULL; + /* No realloc should need this, but just in case... */ + if (!ptr) + return malloc(size * num); + return realloc(ptr, size * num); +}