CAUTION: This patch compiles, but is otherwise totally untested! This patch also implements --times-only. Implementation details for the --source-filter and -dest-filter options: - These options open a *HUGE* security hole in daemon mode unless they are refused in your rsyncd.conf! - Filtering disables rsync alogrithm. (This should be fixed.) - Source filter makes temporary files in /tmp. (Should be overridable.) - If source filter fails, data is send unfiltered. (Should be changed to abort.) - Failure of destination filter, causes data loss!!! (Should be changed to abort.) - If filter changes size of file, you should use --times-only option to prevent repeated transfers of unchanged files. - If the COMMAND contains single quotes, option-passing breaks. (Needs to be fixed.) You should run "make proto" before running "make". --- orig/generator.c 2006-01-20 21:12:17 +++ generator.c 2005-08-17 07:28:01 @@ -59,6 +59,7 @@ extern int append_mode; extern int make_backups; extern int csum_length; extern int ignore_times; +extern int times_only; extern int size_only; extern OFF_T max_size; extern OFF_T min_size; @@ -378,7 +379,7 @@ void itemize(struct file_struct *file, i /* Perform our quick-check heuristic for determining if a file is unchanged. */ int unchanged_file(char *fn, struct file_struct *file, STRUCT_STAT *st) { - if (st->st_size != file->length) + if (!times_only && st->st_size != file->length) return 0; /* if always checksum is set then we use the checksum instead --- orig/options.c 2006-01-21 07:55:00 +++ options.c 2006-01-21 08:09:48 @@ -93,6 +93,7 @@ int keep_partial = 0; int safe_symlinks = 0; int copy_unsafe_links = 0; int size_only = 0; +int times_only = 0; int daemon_bwlimit = 0; int bwlimit = 0; int fuzzy_basis = 0; @@ -142,6 +143,8 @@ char *basis_dir[MAX_BASIS_DIRS+1]; char *config_file = NULL; char *shell_cmd = NULL; char *log_format = NULL; +char *source_filter = NULL; +char *dest_filter = NULL; char *password_file = NULL; char *rsync_path = RSYNC_PATH; char *backup_dir = NULL; @@ -321,6 +324,7 @@ void usage(enum logcode F) rprintf(F," --timeout=TIME set I/O timeout in seconds\n"); rprintf(F," -I, --ignore-times don't skip files that match in size and mod-time\n"); rprintf(F," --size-only skip files that match in size\n"); + rprintf(F," --times-only skip files that match in mod-time\n"); rprintf(F," --modify-window=NUM compare mod-times with reduced accuracy\n"); rprintf(F," -T, --temp-dir=DIR create temporary files in directory DIR\n"); rprintf(F," -y, --fuzzy find similar file for basis if no dest file\n"); @@ -355,6 +359,8 @@ void usage(enum logcode F) rprintf(F," --write-batch=FILE write a batched update to FILE\n"); rprintf(F," --only-write-batch=FILE like --write-batch but w/o updating destination\n"); rprintf(F," --read-batch=FILE read a batched update from FILE\n"); + rprintf(F," --source-filter=COMMAND filter file through COMMAND at source\n"); + rprintf(F," --dest-filter=COMMAND filter file through COMMAND at destination\n"); rprintf(F," --protocol=NUM force an older protocol version to be used\n"); #ifdef INET6 rprintf(F," -4, --ipv4 prefer IPv4\n"); @@ -425,6 +431,7 @@ static struct poptOption long_options[] {"chmod", 0, POPT_ARG_STRING, &chmod_mode, 0, 0, 0 }, {"ignore-times", 'I', POPT_ARG_NONE, &ignore_times, 0, 0, 0 }, {"size-only", 0, POPT_ARG_NONE, &size_only, 0, 0, 0 }, + {"times-only", 0, POPT_ARG_NONE, ×_only , 0, 0, 0 }, {"one-file-system", 'x', POPT_ARG_NONE, 0, 'x', 0, 0 }, {"update", 'u', POPT_ARG_NONE, &update_only, 0, 0, 0 }, {"existing", 0, POPT_ARG_NONE, &ignore_non_existing, 0, 0, 0 }, @@ -496,6 +503,8 @@ static struct poptOption long_options[] {"password-file", 0, POPT_ARG_STRING, &password_file, 0, 0, 0 }, {"blocking-io", 0, POPT_ARG_VAL, &blocking_io, 1, 0, 0 }, {"no-blocking-io", 0, POPT_ARG_VAL, &blocking_io, 0, 0, 0 }, + {"source-filter", 0, POPT_ARG_STRING, &source_filter, 0, 0, 0 }, + {"dest-filter", 0, POPT_ARG_STRING, &dest_filter, 0, 0, 0 }, {"protocol", 0, POPT_ARG_INT, &protocol_version, 0, 0, 0 }, {"checksum-seed", 0, POPT_ARG_INT, &checksum_seed, 0, 0, 0 }, {"server", 0, POPT_ARG_NONE, &am_server, 0, 0, 0 }, @@ -1331,6 +1340,16 @@ int parse_arguments(int *argc, const cha } } + if (source_filter || dest_filter) { + if (whole_file == 0) { + snprintf(err_buf, sizeof err_buf, + "--no-whole-file cannot be used with --%s-filter\n", + source_filter ? "source" : "dest"); + return 0; + } + whole_file = 1; + } + if (files_from) { char *h, *p; int q; @@ -1570,6 +1589,25 @@ void server_options(char **args,int *arg args[ac++] = "--only-write-batch=X"; } + if (times_only && am_sender) + args[ac++] = "--times-only"; + + if (source_filter && !am_sender) { + /* Need to single quote the arg to keep the remote shell + * from splitting it. FIXME: breaks if command has single quotes. */ + if (asprintf(&arg, "--source-filter='%s'", source_filter) < 0) + goto oom; + args[ac++] = arg; + } + + if (dest_filter && am_sender) { + /* Need to single quote the arg to keep the remote shell + * from splitting it. FIXME: breaks if command has single quotes. */ + if (asprintf(&arg, "--dest-filter='%s'", dest_filter) < 0) + goto oom; + args[ac++] = arg; + } + if (size_only) args[ac++] = "--size-only"; --- orig/pipe.c 2006-01-21 08:03:40 +++ pipe.c 2006-01-14 08:34:59 @@ -157,3 +157,77 @@ pid_t local_child(int argc, char **argv, return pid; } + +pid_t run_filter(char *command[], int out, int *pipe_to_filter) +{ + pid_t pid; + int pipefds[2]; + + if (verbose >= 2) + print_child_argv(command); + + if (pipe(pipefds) < 0) { + rsyserr(FERROR, errno, "pipe"); + exit_cleanup(RERR_IPC); + } + + pid = do_fork(); + if (pid == -1) { + rsyserr(FERROR, errno, "fork"); + exit_cleanup(RERR_IPC); + } + + if (pid == 0) { + if (dup2(pipefds[0], STDIN_FILENO) < 0 + || close(pipefds[1]) < 0 + || dup2(out, STDOUT_FILENO) < 0) { + rsyserr(FERROR, errno, "Failed dup/close"); + exit_cleanup(RERR_IPC); + } + umask(orig_umask); + set_blocking(STDIN_FILENO); + if (blocking_io) + set_blocking(STDOUT_FILENO); + execvp(command[0], command); + rsyserr(FERROR, errno, "Failed to exec %s", command[0]); + exit_cleanup(RERR_IPC); + } + + if (close(pipefds[0]) < 0) { + rsyserr(FERROR, errno, "Failed to close"); + exit_cleanup(RERR_IPC); + } + + *pipe_to_filter = pipefds[1]; + + return pid; +} + +pid_t run_filter_on_file(char *command[], int out, int in) +{ + pid_t pid; + + if (verbose >= 2) + print_child_argv(command); + + pid = do_fork(); + if (pid == -1) { + rsyserr(FERROR, errno, "fork"); + exit_cleanup(RERR_IPC); + } + + if (pid == 0) { + if (dup2(in, STDIN_FILENO) < 0 + || dup2(out, STDOUT_FILENO) < 0) { + rsyserr(FERROR, errno, "Failed to dup2"); + exit_cleanup(RERR_IPC); + } + if (blocking_io) + set_blocking(STDOUT_FILENO); + execvp(command[0], command); + rsyserr(FERROR, errno, "Failed to exec %s", command[0]); + exit_cleanup(RERR_IPC); + } + + return pid; +} --- orig/receiver.c 2006-01-14 20:27:09 +++ receiver.c 2005-08-17 07:57:33 @@ -53,6 +53,7 @@ extern int inplace; extern int delay_updates; extern struct stats stats; extern char *log_format; +extern char *dest_filter; extern char *tmpdir; extern char *partial_dir; extern char *basis_dir[]; @@ -411,6 +412,8 @@ int recv_files(int f_in, struct file_lis : !am_server && log_format_has_i; int max_phase = protocol_version >= 29 ? 2 : 1; int i, recv_ok; + pid_t pid = 0; + char *filter_argv[MAX_FILTER_ARGS + 1]; if (verbose > 2) rprintf(FINFO,"recv_files(%d) starting\n",flist->count); @@ -423,6 +426,23 @@ int recv_files(int f_in, struct file_lis if (delay_updates) init_delayed_bits(flist->count); + if (dest_filter) { + char *p; + char *sep = " \t"; + int i; + for (p = strtok(dest_filter, sep), i = 0; + p && i < MAX_FILTER_ARGS; + p = strtok(0, sep)) + filter_argv[i++] = p; + filter_argv[i] = NULL; + if (p) { + rprintf(FERROR, + "Too many arguments to dest-filter (> %d)\n", + MAX_FILTER_ARGS); + exit_cleanup(RERR_SYNTAX); + } + } + while (1) { cleanup_disable(); @@ -665,6 +685,9 @@ int recv_files(int f_in, struct file_lis else if (!am_server && verbose && do_progress) rprintf(FINFO, "%s\n", fname); + if (dest_filter) + pid = run_filter(filter_argv, fd2, &fd2); + /* recv file data */ recv_ok = receive_data(f_in, fnamecmp, fd1, st.st_size, fname, fd2, file->length); @@ -680,6 +703,16 @@ int recv_files(int f_in, struct file_lis exit_cleanup(RERR_FILEIO); } + if (dest_filter) { + int status; + wait_process(pid, &status); + if (status != 0) { + rprintf(FERROR, "filter %s exited code: %d\n", + dest_filter, status); + continue; + } + } + if ((recv_ok && (!delay_updates || !partialptr)) || inplace) { finish_transfer(fname, fnametmp, file, recv_ok, 1); if (partialptr != fname && fnamecmp == partialptr) { --- orig/rsync.h 2006-01-14 20:27:10 +++ rsync.h 2005-08-17 07:10:11 @@ -103,6 +103,7 @@ #define IOERR_DEL_LIMIT (1<<2) #define MAX_ARGS 1000 +#define MAX_FILTER_ARGS 100 #define MAX_BASIS_DIRS 20 #define MAX_SERVER_ARGS (MAX_BASIS_DIRS*2 + 100) --- orig/rsync.yo 2006-01-21 08:12:23 +++ rsync.yo 2005-08-17 07:08:21 @@ -350,6 +350,7 @@ to the detailed description below for a --timeout=TIME set I/O timeout in seconds -I, --ignore-times don't skip files that match size and time --size-only skip files that match in size + --times-only skip files that match in mod-time --modify-window=NUM compare mod-times with reduced accuracy -T, --temp-dir=DIR create temporary files in directory DIR -y, --fuzzy find similar file for basis if no dest file @@ -384,6 +385,8 @@ to the detailed description below for a --write-batch=FILE write a batched update to FILE --only-write-batch=FILE like --write-batch but w/o updating dest --read-batch=FILE read a batched update from FILE + --source-filter=COMMAND filter file through COMMAND at source + --dest-filter=COMMAND filter file through COMMAND at destination --protocol=NUM force an older protocol version to be used --checksum-seed=NUM set block/file checksum seed (advanced) -4, --ipv4 prefer IPv4 @@ -1430,6 +1433,33 @@ file previously generated by bf(--write- If em(FILE) is bf(-), the batch data will be read from standard input. See the "BATCH MODE" section for details. +dit(bf(--source-filter=COMMAND)) This option allows the user to specify a +filter program that will be applied to the contents of all transferred +regular files before the data is sent to destination. COMMAND will receive +the data on its standard input and it should write the filtered data to +standard output. COMMAND should exit non-zero if it cannot process the +data or if it encounters an error when writing the data to stdout. + +Example: --source-filter="gzip -9" will cause remote files to be +compressed. +Use of --source-filter automatically enables --whole-file. +If your filter does not output the same number of bytes that it received on +input, you should use --times-only to disable size and content checks on +subsequent rsync runs. + +dit(bf(--dest-filter=COMMAND)) This option allows you to specify a filter +program that will be applied to the contents of all transferred regular +files before the data is written to disk. COMMAND will receive the data on +its standard input and it should write the filtered data to standard +output. COMMAND should exit non-zero if it cannot process the data or if +it encounters an error when writing the data to stdout. + +Example: --dest-filter="gzip -9" will cause remote files to be compressed. +Use of --dest-filter automatically enables --whole-file. +If your filter does not output the same number of bytes that it +received on input, you should use --times-only to disable size and +content checks on subsequent rsync runs. + dit(bf(--protocol=NUM)) Force an older protocol version to be used. This is useful for creating a batch file that is compatible with an older version of rsync. For instance, if rsync 2.6.4 is being used with the --- orig/sender.c 2006-01-14 20:27:10 +++ sender.c 2005-08-17 07:40:49 @@ -41,6 +41,7 @@ extern int write_batch; extern struct stats stats; extern struct file_list *the_file_list; extern char *log_format; +extern char *source_filter; /** @@ -219,6 +220,26 @@ void send_files(struct file_list *flist, : !am_server && log_format_has_i; int f_xfer = write_batch < 0 ? batch_fd : f_out; int i, j; + char *filter_argv[MAX_FILTER_ARGS + 1]; + char *tmp = 0; + int unlink_tmp = 0; + + if (source_filter) { + char *p; + char *sep = " \t"; + int i; + for (p = strtok(source_filter, sep), i = 0; + p && i < MAX_FILTER_ARGS; + p = strtok(0, sep)) + filter_argv[i++] = p; + filter_argv[i] = NULL; + if (p) { + rprintf(FERROR, + "Too many arguments to source-filter (> %d)\n", + MAX_FILTER_ARGS); + exit_cleanup(RERR_SYNTAX); + } + } if (verbose > 2) rprintf(FINFO, "send_files starting\n"); @@ -293,6 +314,7 @@ void send_files(struct file_list *flist, return; } + unlink_tmp = 0; fd = do_open(fname, O_RDONLY, 0); if (fd == -1) { if (errno == ENOENT) { @@ -321,6 +343,33 @@ void send_files(struct file_list *flist, return; } + if (source_filter) { + int fd2; + char *tmpl = "/tmp/rsync-filtered_sourceXXXXXX"; + + tmp = strdup(tmpl); + fd2 = mkstemp(tmp); + if (fd2 == -1) { + rprintf(FERROR, "mkstemp %s failed: %s\n", + tmp, strerror(errno)); + } else { + int status; + pid_t pid = run_filter_on_file(filter_argv, fd2, fd); + close(fd); + close(fd2); + wait_process(pid, &status); + if (status != 0) { + rprintf(FERROR, + "bypassing source filter %s; exited with code: %d\n", + source_filter, status); + fd = do_open(fname, O_RDONLY, 0); + } else { + fd = do_open(tmp, O_RDONLY, 0); + unlink_tmp = 1; + } + } + } + if (st.st_size) { int32 read_size = MAX(s->blength * 3, MAX_MAP_SIZE); mbuf = map_file(fd, st.st_size, read_size, s->blength); @@ -363,6 +412,8 @@ void send_files(struct file_list *flist, } } close(fd); + if (unlink_tmp) + unlink(tmp); free_sums(s);