Annotation of embedaddon/rsync/io.c, revision 1.1.1.1
1.1 misho 1: /*
2: * Socket and pipe I/O utilities used in rsync.
3: *
4: * Copyright (C) 1996-2001 Andrew Tridgell
5: * Copyright (C) 1996 Paul Mackerras
6: * Copyright (C) 2001, 2002 Martin Pool <mbp@samba.org>
7: * Copyright (C) 2003-2009 Wayne Davison
8: *
9: * This program is free software; you can redistribute it and/or modify
10: * it under the terms of the GNU General Public License as published by
11: * the Free Software Foundation; either version 3 of the License, or
12: * (at your option) any later version.
13: *
14: * This program is distributed in the hope that it will be useful,
15: * but WITHOUT ANY WARRANTY; without even the implied warranty of
16: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17: * GNU General Public License for more details.
18: *
19: * You should have received a copy of the GNU General Public License along
20: * with this program; if not, visit the http://fsf.org website.
21: */
22:
23: /* Rsync provides its own multiplexing system, which is used to send
24: * stderr and stdout over a single socket.
25: *
26: * For historical reasons this is off during the start of the
27: * connection, but it's switched on quite early using
28: * io_start_multiplex_out() and io_start_multiplex_in(). */
29:
30: #include "rsync.h"
31: #include "ifuncs.h"
32:
33: /** If no timeout is specified then use a 60 second select timeout */
34: #define SELECT_TIMEOUT 60
35:
36: extern int bwlimit;
37: extern size_t bwlimit_writemax;
38: extern int io_timeout;
39: extern int am_server;
40: extern int am_daemon;
41: extern int am_sender;
42: extern int am_generator;
43: extern int inc_recurse;
44: extern int io_error;
45: extern int eol_nulls;
46: extern int flist_eof;
47: extern int list_only;
48: extern int read_batch;
49: extern int compat_flags;
50: extern int protect_args;
51: extern int checksum_seed;
52: extern int protocol_version;
53: extern int remove_source_files;
54: extern int preserve_hard_links;
55: extern struct stats stats;
56: extern struct file_list *cur_flist;
57: #ifdef ICONV_OPTION
58: extern int filesfrom_convert;
59: extern iconv_t ic_send, ic_recv;
60: #endif
61:
62: int csum_length = SHORT_SUM_LENGTH; /* initial value */
63: int allowed_lull = 0;
64: int ignore_timeout = 0;
65: int batch_fd = -1;
66: int msgdone_cnt = 0;
67:
68: /* Ignore an EOF error if non-zero. See whine_about_eof(). */
69: int kluge_around_eof = 0;
70:
71: int msg_fd_in = -1;
72: int msg_fd_out = -1;
73: int sock_f_in = -1;
74: int sock_f_out = -1;
75:
76: static int iobuf_f_in = -1;
77: static char *iobuf_in;
78: static size_t iobuf_in_siz;
79: static size_t iobuf_in_ndx;
80: static size_t iobuf_in_remaining;
81:
82: static int iobuf_f_out = -1;
83: static char *iobuf_out;
84: static int iobuf_out_cnt;
85:
86: int flist_forward_from = -1;
87:
88: static int io_multiplexing_out;
89: static int io_multiplexing_in;
90: static time_t last_io_in;
91: static time_t last_io_out;
92: static int no_flush;
93:
94: static int write_batch_monitor_in = -1;
95: static int write_batch_monitor_out = -1;
96:
97: static int io_filesfrom_f_in = -1;
98: static int io_filesfrom_f_out = -1;
99: static xbuf ff_buf = EMPTY_XBUF;
100: static char ff_lastchar;
101: #ifdef ICONV_OPTION
102: static xbuf iconv_buf = EMPTY_XBUF;
103: #endif
104: static int defer_forwarding_messages = 0, keep_defer_forwarding = 0;
105: static int select_timeout = SELECT_TIMEOUT;
106: static int active_filecnt = 0;
107: static OFF_T active_bytecnt = 0;
108: static int first_message = 1;
109:
110: static char int_byte_extra[64] = {
111: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* (00 - 3F)/4 */
112: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* (40 - 7F)/4 */
113: 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, /* (80 - BF)/4 */
114: 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 5, 6, /* (C0 - FF)/4 */
115: };
116:
117: #define REMOTE_OPTION_ERROR "rsync: on remote machine: -"
118: #define REMOTE_OPTION_ERROR2 ": unknown option"
119:
120: enum festatus { FES_SUCCESS, FES_REDO, FES_NO_SEND };
121:
122: static void check_timeout(void)
123: {
124: time_t t, chk;
125:
126: if (!io_timeout || ignore_timeout)
127: return;
128:
129: t = time(NULL);
130:
131: if (!last_io_in)
132: last_io_in = t;
133:
134: chk = MAX(last_io_out, last_io_in);
135: if (t - chk >= io_timeout) {
136: if (am_server || am_daemon)
137: exit_cleanup(RERR_TIMEOUT);
138: rprintf(FERROR, "[%s] io timeout after %d seconds -- exiting\n",
139: who_am_i(), (int)(t-chk));
140: exit_cleanup(RERR_TIMEOUT);
141: }
142: }
143:
144: static void readfd(int fd, char *buffer, size_t N);
145: static void writefd(int fd, const char *buf, size_t len);
146: static void writefd_unbuffered(int fd, const char *buf, size_t len);
147: static void mplex_write(int fd, enum msgcode code, const char *buf, size_t len, int convert);
148:
149: static flist_ndx_list redo_list, hlink_list;
150:
151: struct msg_list_item {
152: struct msg_list_item *next;
153: char convert;
154: char buf[1];
155: };
156:
157: struct msg_list {
158: struct msg_list_item *head, *tail;
159: };
160:
161: static struct msg_list msg_queue;
162:
163: static void got_flist_entry_status(enum festatus status, const char *buf)
164: {
165: int ndx = IVAL(buf, 0);
166: struct file_list *flist = flist_for_ndx(ndx, "got_flist_entry_status");
167:
168: if (remove_source_files) {
169: active_filecnt--;
170: active_bytecnt -= F_LENGTH(flist->files[ndx - flist->ndx_start]);
171: }
172:
173: if (inc_recurse)
174: flist->in_progress--;
175:
176: switch (status) {
177: case FES_SUCCESS:
178: if (remove_source_files)
179: send_msg(MSG_SUCCESS, buf, 4, 0);
180: /* FALL THROUGH */
181: case FES_NO_SEND:
182: #ifdef SUPPORT_HARD_LINKS
183: if (preserve_hard_links) {
184: struct file_struct *file = flist->files[ndx - flist->ndx_start];
185: if (F_IS_HLINKED(file)) {
186: if (status == FES_NO_SEND)
187: flist_ndx_push(&hlink_list, -2); /* indicates a failure follows */
188: flist_ndx_push(&hlink_list, ndx);
189: flist->in_progress++;
190: }
191: }
192: #endif
193: break;
194: case FES_REDO:
195: if (read_batch) {
196: if (inc_recurse)
197: flist->in_progress++;
198: break;
199: }
200: if (inc_recurse)
201: flist->to_redo++;
202: flist_ndx_push(&redo_list, ndx);
203: break;
204: }
205: }
206:
207: /* Note the fds used for the main socket (which might really be a pipe
208: * for a local transfer, but we can ignore that). */
209: void io_set_sock_fds(int f_in, int f_out)
210: {
211: sock_f_in = f_in;
212: sock_f_out = f_out;
213: }
214:
215: void set_io_timeout(int secs)
216: {
217: io_timeout = secs;
218: allowed_lull = (io_timeout + 1) / 2;
219:
220: if (!io_timeout || allowed_lull > SELECT_TIMEOUT)
221: select_timeout = SELECT_TIMEOUT;
222: else
223: select_timeout = allowed_lull;
224:
225: if (read_batch)
226: allowed_lull = 0;
227: }
228:
229: /* Setup the fd used to receive MSG_* messages. Only needed during the
230: * early stages of being a local sender (up through the sending of the
231: * file list) or when we're the generator (to fetch the messages from
232: * the receiver). */
233: void set_msg_fd_in(int fd)
234: {
235: msg_fd_in = fd;
236: }
237:
238: /* Setup the fd used to send our MSG_* messages. Only needed when
239: * we're the receiver (to send our messages to the generator). */
240: void set_msg_fd_out(int fd)
241: {
242: msg_fd_out = fd;
243: set_nonblocking(msg_fd_out);
244: }
245:
246: /* Add a message to the pending MSG_* list. */
247: static void msg_list_add(struct msg_list *lst, int code, const char *buf, int len, int convert)
248: {
249: struct msg_list_item *m;
250: int sz = len + 4 + sizeof m[0] - 1;
251:
252: if (!(m = (struct msg_list_item *)new_array(char, sz)))
253: out_of_memory("msg_list_add");
254: m->next = NULL;
255: m->convert = convert;
256: SIVAL(m->buf, 0, ((code+MPLEX_BASE)<<24) | len);
257: memcpy(m->buf + 4, buf, len);
258: if (lst->tail)
259: lst->tail->next = m;
260: else
261: lst->head = m;
262: lst->tail = m;
263: }
264:
265: static inline int flush_a_msg(int fd)
266: {
267: struct msg_list_item *m = msg_queue.head;
268: int len = IVAL(m->buf, 0) & 0xFFFFFF;
269: int tag = *((uchar*)m->buf+3) - MPLEX_BASE;
270:
271: if (!(msg_queue.head = m->next))
272: msg_queue.tail = NULL;
273:
274: defer_forwarding_messages++;
275: mplex_write(fd, tag, m->buf + 4, len, m->convert);
276: defer_forwarding_messages--;
277:
278: free(m);
279:
280: return len;
281: }
282:
283: static void msg_flush(void)
284: {
285: if (am_generator) {
286: while (msg_queue.head && io_multiplexing_out)
287: stats.total_written += flush_a_msg(sock_f_out) + 4;
288: } else {
289: while (msg_queue.head)
290: (void)flush_a_msg(msg_fd_out);
291: }
292: }
293:
294: static void check_for_d_option_error(const char *msg)
295: {
296: static char rsync263_opts[] = "BCDHIKLPRSTWabceghlnopqrtuvxz";
297: char *colon;
298: int saw_d = 0;
299:
300: if (*msg != 'r'
301: || strncmp(msg, REMOTE_OPTION_ERROR, sizeof REMOTE_OPTION_ERROR - 1) != 0)
302: return;
303:
304: msg += sizeof REMOTE_OPTION_ERROR - 1;
305: if (*msg == '-' || (colon = strchr(msg, ':')) == NULL
306: || strncmp(colon, REMOTE_OPTION_ERROR2, sizeof REMOTE_OPTION_ERROR2 - 1) != 0)
307: return;
308:
309: for ( ; *msg != ':'; msg++) {
310: if (*msg == 'd')
311: saw_d = 1;
312: else if (*msg == 'e')
313: break;
314: else if (strchr(rsync263_opts, *msg) == NULL)
315: return;
316: }
317:
318: if (saw_d) {
319: rprintf(FWARNING,
320: "*** Try using \"--old-d\" if remote rsync is <= 2.6.3 ***\n");
321: }
322: }
323:
324: /* Read a message from the MSG_* fd and handle it. This is called either
325: * during the early stages of being a local sender (up through the sending
326: * of the file list) or when we're the generator (to fetch the messages
327: * from the receiver). */
328: static void read_msg_fd(void)
329: {
330: char buf[2048];
331: size_t n;
332: struct file_list *flist;
333: int fd = msg_fd_in;
334: int tag, len;
335:
336: /* Temporarily disable msg_fd_in. This is needed to avoid looping back
337: * to this routine from writefd_unbuffered(). */
338: no_flush++;
339: msg_fd_in = -1;
340: defer_forwarding_messages++;
341:
342: readfd(fd, buf, 4);
343: tag = IVAL(buf, 0);
344:
345: len = tag & 0xFFFFFF;
346: tag = (tag >> 24) - MPLEX_BASE;
347:
348: switch (tag) {
349: case MSG_DONE:
350: if (len < 0 || len > 1 || !am_generator) {
351: invalid_msg:
352: rprintf(FERROR, "invalid message %d:%d [%s%s]\n",
353: tag, len, who_am_i(),
354: inc_recurse ? "/inc" : "");
355: exit_cleanup(RERR_STREAMIO);
356: }
357: if (len) {
358: readfd(fd, buf, len);
359: stats.total_read = read_varlong(fd, 3);
360: }
361: msgdone_cnt++;
362: break;
363: case MSG_REDO:
364: if (len != 4 || !am_generator)
365: goto invalid_msg;
366: readfd(fd, buf, 4);
367: got_flist_entry_status(FES_REDO, buf);
368: break;
369: case MSG_FLIST:
370: if (len != 4 || !am_generator || !inc_recurse)
371: goto invalid_msg;
372: readfd(fd, buf, 4);
373: /* Read extra file list from receiver. */
374: assert(iobuf_in != NULL);
375: assert(iobuf_f_in == fd);
376: if (verbose > 3) {
377: rprintf(FINFO, "[%s] receiving flist for dir %d\n",
378: who_am_i(), IVAL(buf,0));
379: }
380: flist = recv_file_list(fd);
381: flist->parent_ndx = IVAL(buf,0);
382: #ifdef SUPPORT_HARD_LINKS
383: if (preserve_hard_links)
384: match_hard_links(flist);
385: #endif
386: break;
387: case MSG_FLIST_EOF:
388: if (len != 0 || !am_generator || !inc_recurse)
389: goto invalid_msg;
390: flist_eof = 1;
391: break;
392: case MSG_IO_ERROR:
393: if (len != 4)
394: goto invalid_msg;
395: readfd(fd, buf, len);
396: io_error |= IVAL(buf, 0);
397: break;
398: case MSG_DELETED:
399: if (len >= (int)sizeof buf || !am_generator)
400: goto invalid_msg;
401: readfd(fd, buf, len);
402: send_msg(MSG_DELETED, buf, len, 1);
403: break;
404: case MSG_SUCCESS:
405: if (len != 4 || !am_generator)
406: goto invalid_msg;
407: readfd(fd, buf, 4);
408: got_flist_entry_status(FES_SUCCESS, buf);
409: break;
410: case MSG_NO_SEND:
411: if (len != 4 || !am_generator)
412: goto invalid_msg;
413: readfd(fd, buf, 4);
414: got_flist_entry_status(FES_NO_SEND, buf);
415: break;
416: case MSG_ERROR_SOCKET:
417: case MSG_ERROR_UTF8:
418: case MSG_CLIENT:
419: if (!am_generator)
420: goto invalid_msg;
421: if (tag == MSG_ERROR_SOCKET)
422: io_end_multiplex_out();
423: /* FALL THROUGH */
424: case MSG_INFO:
425: case MSG_ERROR:
426: case MSG_ERROR_XFER:
427: case MSG_WARNING:
428: case MSG_LOG:
429: while (len) {
430: n = len;
431: if (n >= sizeof buf)
432: n = sizeof buf - 1;
433: readfd(fd, buf, n);
434: rwrite((enum logcode)tag, buf, n, !am_generator);
435: len -= n;
436: }
437: break;
438: default:
439: rprintf(FERROR, "unknown message %d:%d [%s]\n",
440: tag, len, who_am_i());
441: exit_cleanup(RERR_STREAMIO);
442: }
443:
444: no_flush--;
445: msg_fd_in = fd;
446: if (!--defer_forwarding_messages && !no_flush)
447: msg_flush();
448: }
449:
450: /* This is used by the generator to limit how many file transfers can
451: * be active at once when --remove-source-files is specified. Without
452: * this, sender-side deletions were mostly happening at the end. */
453: void increment_active_files(int ndx, int itemizing, enum logcode code)
454: {
455: while (1) {
456: /* TODO: tune these limits? */
457: int limit = active_bytecnt >= 128*1024 ? 10 : 50;
458: if (active_filecnt < limit)
459: break;
460: check_for_finished_files(itemizing, code, 0);
461: if (active_filecnt < limit)
462: break;
463: if (iobuf_out_cnt)
464: io_flush(NORMAL_FLUSH);
465: else
466: read_msg_fd();
467: }
468:
469: active_filecnt++;
470: active_bytecnt += F_LENGTH(cur_flist->files[ndx - cur_flist->ndx_start]);
471: }
472:
473: /* Write an message to a multiplexed stream. If this fails, rsync exits. */
474: static void mplex_write(int fd, enum msgcode code, const char *buf, size_t len, int convert)
475: {
476: char buffer[BIGPATHBUFLEN]; /* Oversized for use by iconv code. */
477: size_t n = len;
478:
479: #ifdef ICONV_OPTION
480: /* We need to convert buf before doing anything else so that we
481: * can include the (converted) byte length in the message header. */
482: if (convert && ic_send != (iconv_t)-1) {
483: xbuf outbuf, inbuf;
484:
485: INIT_XBUF(outbuf, buffer + 4, 0, sizeof buffer - 4);
486: INIT_XBUF(inbuf, (char*)buf, len, -1);
487:
488: iconvbufs(ic_send, &inbuf, &outbuf,
489: ICB_INCLUDE_BAD | ICB_INCLUDE_INCOMPLETE);
490: if (inbuf.len > 0) {
491: rprintf(FERROR, "overflowed conversion buffer in mplex_write");
492: exit_cleanup(RERR_UNSUPPORTED);
493: }
494:
495: n = len = outbuf.len;
496: } else
497: #endif
498: if (n > 1024 - 4) /* BIGPATHBUFLEN can handle 1024 bytes */
499: n = 0; /* We'd rather do 2 writes than too much memcpy(). */
500: else
501: memcpy(buffer + 4, buf, n);
502:
503: SIVAL(buffer, 0, ((MPLEX_BASE + (int)code)<<24) + len);
504:
505: keep_defer_forwarding++; /* defer_forwarding_messages++ on return */
506: writefd_unbuffered(fd, buffer, n+4);
507: keep_defer_forwarding--;
508:
509: if (len > n)
510: writefd_unbuffered(fd, buf+n, len-n);
511:
512: if (!--defer_forwarding_messages && !no_flush)
513: msg_flush();
514: }
515:
516: int send_msg(enum msgcode code, const char *buf, int len, int convert)
517: {
518: if (msg_fd_out < 0) {
519: if (!defer_forwarding_messages)
520: return io_multiplex_write(code, buf, len, convert);
521: if (!io_multiplexing_out)
522: return 0;
523: msg_list_add(&msg_queue, code, buf, len, convert);
524: return 1;
525: }
526: if (flist_forward_from >= 0)
527: msg_list_add(&msg_queue, code, buf, len, convert);
528: else
529: mplex_write(msg_fd_out, code, buf, len, convert);
530: return 1;
531: }
532:
533: void send_msg_int(enum msgcode code, int num)
534: {
535: char numbuf[4];
536: SIVAL(numbuf, 0, num);
537: send_msg(code, numbuf, 4, 0);
538: }
539:
540: void wait_for_receiver(void)
541: {
542: if (io_flush(NORMAL_FLUSH))
543: return;
544: read_msg_fd();
545: }
546:
547: int get_redo_num(void)
548: {
549: return flist_ndx_pop(&redo_list);
550: }
551:
552: int get_hlink_num(void)
553: {
554: return flist_ndx_pop(&hlink_list);
555: }
556:
557: /**
558: * When we're the receiver and we have a local --files-from list of names
559: * that needs to be sent over the socket to the sender, we have to do two
560: * things at the same time: send the sender a list of what files we're
561: * processing and read the incoming file+info list from the sender. We do
562: * this by augmenting the read_timeout() function to copy this data. It
563: * uses ff_buf to read a block of data from f_in (when it is ready, since
564: * it might be a pipe) and then blast it out f_out (when it is ready to
565: * receive more data).
566: */
567: void io_set_filesfrom_fds(int f_in, int f_out)
568: {
569: io_filesfrom_f_in = f_in;
570: io_filesfrom_f_out = f_out;
571: alloc_xbuf(&ff_buf, 2048);
572: #ifdef ICONV_OPTION
573: if (protect_args)
574: alloc_xbuf(&iconv_buf, 1024);
575: #endif
576: }
577:
578: /* It's almost always an error to get an EOF when we're trying to read from the
579: * network, because the protocol is (for the most part) self-terminating.
580: *
581: * There is one case for the receiver when it is at the end of the transfer
582: * (hanging around reading any keep-alive packets that might come its way): if
583: * the sender dies before the generator's kill-signal comes through, we can end
584: * up here needing to loop until the kill-signal arrives. In this situation,
585: * kluge_around_eof will be < 0.
586: *
587: * There is another case for older protocol versions (< 24) where the module
588: * listing was not terminated, so we must ignore an EOF error in that case and
589: * exit. In this situation, kluge_around_eof will be > 0. */
590: static void whine_about_eof(int fd)
591: {
592: if (kluge_around_eof && fd == sock_f_in) {
593: int i;
594: if (kluge_around_eof > 0)
595: exit_cleanup(0);
596: /* If we're still here after 10 seconds, exit with an error. */
597: for (i = 10*1000/20; i--; )
598: msleep(20);
599: }
600:
601: rprintf(FERROR, RSYNC_NAME ": connection unexpectedly closed "
602: "(%.0f bytes received so far) [%s]\n",
603: (double)stats.total_read, who_am_i());
604:
605: exit_cleanup(RERR_STREAMIO);
606: }
607:
608: /**
609: * Read from a socket with I/O timeout. return the number of bytes
610: * read. If no bytes can be read then exit, never return a number <= 0.
611: *
612: * TODO: If the remote shell connection fails, then current versions
613: * actually report an "unexpected EOF" error here. Since it's a
614: * fairly common mistake to try to use rsh when ssh is required, we
615: * should trap that: if we fail to read any data at all, we should
616: * give a better explanation. We can tell whether the connection has
617: * started by looking e.g. at whether the remote version is known yet.
618: */
619: static int read_timeout(int fd, char *buf, size_t len)
620: {
621: int n, cnt = 0;
622:
623: io_flush(FULL_FLUSH);
624:
625: while (cnt == 0) {
626: /* until we manage to read *something* */
627: fd_set r_fds, w_fds;
628: struct timeval tv;
629: int maxfd = fd;
630: int count;
631:
632: FD_ZERO(&r_fds);
633: FD_ZERO(&w_fds);
634: FD_SET(fd, &r_fds);
635: if (io_filesfrom_f_out >= 0) {
636: int new_fd;
637: if (ff_buf.len == 0) {
638: if (io_filesfrom_f_in >= 0) {
639: FD_SET(io_filesfrom_f_in, &r_fds);
640: new_fd = io_filesfrom_f_in;
641: } else {
642: io_filesfrom_f_out = -1;
643: new_fd = -1;
644: }
645: } else {
646: FD_SET(io_filesfrom_f_out, &w_fds);
647: new_fd = io_filesfrom_f_out;
648: }
649: if (new_fd > maxfd)
650: maxfd = new_fd;
651: }
652:
653: tv.tv_sec = select_timeout;
654: tv.tv_usec = 0;
655:
656: errno = 0;
657:
658: count = select(maxfd + 1, &r_fds, &w_fds, NULL, &tv);
659:
660: if (count <= 0) {
661: if (errno == EBADF) {
662: defer_forwarding_messages = 0;
663: exit_cleanup(RERR_SOCKETIO);
664: }
665: check_timeout();
666: continue;
667: }
668:
669: if (io_filesfrom_f_out >= 0) {
670: if (ff_buf.len) {
671: if (FD_ISSET(io_filesfrom_f_out, &w_fds)) {
672: int l = write(io_filesfrom_f_out,
673: ff_buf.buf + ff_buf.pos,
674: ff_buf.len);
675: if (l > 0) {
676: if (!(ff_buf.len -= l))
677: ff_buf.pos = 0;
678: else
679: ff_buf.pos += l;
680: } else if (errno != EINTR) {
681: /* XXX should we complain? */
682: io_filesfrom_f_out = -1;
683: }
684: }
685: } else if (io_filesfrom_f_in >= 0) {
686: if (FD_ISSET(io_filesfrom_f_in, &r_fds)) {
687: #ifdef ICONV_OPTION
688: xbuf *ibuf = filesfrom_convert ? &iconv_buf : &ff_buf;
689: #else
690: xbuf *ibuf = &ff_buf;
691: #endif
692: int l = read(io_filesfrom_f_in, ibuf->buf, ibuf->size);
693: if (l <= 0) {
694: if (l == 0 || errno != EINTR) {
695: /* Send end-of-file marker */
696: memcpy(ff_buf.buf, "\0\0", 2);
697: ff_buf.len = ff_lastchar? 2 : 1;
698: ff_buf.pos = 0;
699: io_filesfrom_f_in = -1;
700: }
701: } else {
702: #ifdef ICONV_OPTION
703: if (filesfrom_convert) {
704: iconv_buf.pos = 0;
705: iconv_buf.len = l;
706: iconvbufs(ic_send, &iconv_buf, &ff_buf,
707: ICB_EXPAND_OUT|ICB_INCLUDE_BAD|ICB_INCLUDE_INCOMPLETE);
708: l = ff_buf.len;
709: }
710: #endif
711: if (!eol_nulls) {
712: char *s = ff_buf.buf + l;
713: /* Transform CR and/or LF into '\0' */
714: while (s-- > ff_buf.buf) {
715: if (*s == '\n' || *s == '\r')
716: *s = '\0';
717: }
718: }
719: if (!ff_lastchar) {
720: /* Last buf ended with a '\0', so don't
721: * let this buf start with one. */
722: while (l && ff_buf.buf[ff_buf.pos] == '\0')
723: ff_buf.pos++, l--;
724: }
725: if (!l)
726: ff_buf.pos = 0;
727: else {
728: char *f = ff_buf.buf + ff_buf.pos;
729: char *t = f;
730: char *eob = f + l;
731: /* Eliminate any multi-'\0' runs. */
732: while (f != eob) {
733: if (!(*t++ = *f++)) {
734: while (f != eob && !*f)
735: f++, l--;
736: }
737: }
738: ff_lastchar = f[-1];
739: }
740: ff_buf.len = l;
741: }
742: }
743: }
744: }
745:
746: if (!FD_ISSET(fd, &r_fds))
747: continue;
748:
749: n = read(fd, buf, len);
750:
751: if (n <= 0) {
752: if (n == 0)
753: whine_about_eof(fd); /* Doesn't return. */
754: if (errno == EINTR || errno == EWOULDBLOCK
755: || errno == EAGAIN)
756: continue;
757:
758: /* Don't write errors on a dead socket. */
759: if (fd == sock_f_in) {
760: io_end_multiplex_out();
761: rsyserr(FERROR_SOCKET, errno, "read error");
762: } else
763: rsyserr(FERROR, errno, "read error");
764: exit_cleanup(RERR_STREAMIO);
765: }
766:
767: buf += n;
768: len -= n;
769: cnt += n;
770:
771: if (fd == sock_f_in && io_timeout)
772: last_io_in = time(NULL);
773: }
774:
775: return cnt;
776: }
777:
778: /* Read a line into the "buf" buffer. */
779: int read_line(int fd, char *buf, size_t bufsiz, int flags)
780: {
781: char ch, *s, *eob;
782: int cnt;
783:
784: #ifdef ICONV_OPTION
785: if (flags & RL_CONVERT && iconv_buf.size < bufsiz)
786: realloc_xbuf(&iconv_buf, bufsiz + 1024);
787: #endif
788:
789: start:
790: #ifdef ICONV_OPTION
791: s = flags & RL_CONVERT ? iconv_buf.buf : buf;
792: #else
793: s = buf;
794: #endif
795: eob = s + bufsiz - 1;
796: while (1) {
797: cnt = read(fd, &ch, 1);
798: if (cnt < 0 && (errno == EWOULDBLOCK
799: || errno == EINTR || errno == EAGAIN)) {
800: struct timeval tv;
801: fd_set r_fds, e_fds;
802: FD_ZERO(&r_fds);
803: FD_SET(fd, &r_fds);
804: FD_ZERO(&e_fds);
805: FD_SET(fd, &e_fds);
806: tv.tv_sec = select_timeout;
807: tv.tv_usec = 0;
808: if (!select(fd+1, &r_fds, NULL, &e_fds, &tv))
809: check_timeout();
810: /*if (FD_ISSET(fd, &e_fds))
811: rprintf(FINFO, "select exception on fd %d\n", fd); */
812: continue;
813: }
814: if (cnt != 1)
815: break;
816: if (flags & RL_EOL_NULLS ? ch == '\0' : (ch == '\r' || ch == '\n')) {
817: /* Skip empty lines if dumping comments. */
818: if (flags & RL_DUMP_COMMENTS && s == buf)
819: continue;
820: break;
821: }
822: if (s < eob)
823: *s++ = ch;
824: }
825: *s = '\0';
826:
827: if (flags & RL_DUMP_COMMENTS && (*buf == '#' || *buf == ';'))
828: goto start;
829:
830: #ifdef ICONV_OPTION
831: if (flags & RL_CONVERT) {
832: xbuf outbuf;
833: INIT_XBUF(outbuf, buf, 0, bufsiz);
834: iconv_buf.pos = 0;
835: iconv_buf.len = s - iconv_buf.buf;
836: iconvbufs(ic_recv, &iconv_buf, &outbuf,
837: ICB_INCLUDE_BAD | ICB_INCLUDE_INCOMPLETE);
838: outbuf.buf[outbuf.len] = '\0';
839: return outbuf.len;
840: }
841: #endif
842:
843: return s - buf;
844: }
845:
846: void read_args(int f_in, char *mod_name, char *buf, size_t bufsiz, int rl_nulls,
847: char ***argv_p, int *argc_p, char **request_p)
848: {
849: int maxargs = MAX_ARGS;
850: int dot_pos = 0;
851: int argc = 0;
852: char **argv, *p;
853: int rl_flags = (rl_nulls ? RL_EOL_NULLS : 0);
854:
855: #ifdef ICONV_OPTION
856: rl_flags |= (protect_args && ic_recv != (iconv_t)-1 ? RL_CONVERT : 0);
857: #endif
858:
859: if (!(argv = new_array(char *, maxargs)))
860: out_of_memory("read_args");
861: if (mod_name && !protect_args)
862: argv[argc++] = "rsyncd";
863:
864: while (1) {
865: if (read_line(f_in, buf, bufsiz, rl_flags) == 0)
866: break;
867:
868: if (argc == maxargs-1) {
869: maxargs += MAX_ARGS;
870: if (!(argv = realloc_array(argv, char *, maxargs)))
871: out_of_memory("read_args");
872: }
873:
874: if (dot_pos) {
875: if (request_p) {
876: *request_p = strdup(buf);
877: request_p = NULL;
878: }
879: if (mod_name)
880: glob_expand_module(mod_name, buf, &argv, &argc, &maxargs);
881: else
882: glob_expand(buf, &argv, &argc, &maxargs);
883: } else {
884: if (!(p = strdup(buf)))
885: out_of_memory("read_args");
886: argv[argc++] = p;
887: if (*p == '.' && p[1] == '\0')
888: dot_pos = argc;
889: }
890: }
891: argv[argc] = NULL;
892:
893: glob_expand(NULL, NULL, NULL, NULL);
894:
895: *argc_p = argc;
896: *argv_p = argv;
897: }
898:
899: int io_start_buffering_out(int f_out)
900: {
901: if (iobuf_out) {
902: assert(f_out == iobuf_f_out);
903: return 0;
904: }
905: if (!(iobuf_out = new_array(char, IO_BUFFER_SIZE)))
906: out_of_memory("io_start_buffering_out");
907: iobuf_out_cnt = 0;
908: iobuf_f_out = f_out;
909: return 1;
910: }
911:
912: int io_start_buffering_in(int f_in)
913: {
914: if (iobuf_in) {
915: assert(f_in == iobuf_f_in);
916: return 0;
917: }
918: iobuf_in_siz = 2 * IO_BUFFER_SIZE;
919: if (!(iobuf_in = new_array(char, iobuf_in_siz)))
920: out_of_memory("io_start_buffering_in");
921: iobuf_f_in = f_in;
922: return 1;
923: }
924:
925: void io_end_buffering_in(void)
926: {
927: if (!iobuf_in)
928: return;
929: free(iobuf_in);
930: iobuf_in = NULL;
931: iobuf_in_ndx = 0;
932: iobuf_in_remaining = 0;
933: iobuf_f_in = -1;
934: }
935:
936: void io_end_buffering_out(void)
937: {
938: if (!iobuf_out)
939: return;
940: io_flush(FULL_FLUSH);
941: free(iobuf_out);
942: iobuf_out = NULL;
943: iobuf_f_out = -1;
944: }
945:
946: void maybe_flush_socket(int important)
947: {
948: if (iobuf_out && iobuf_out_cnt
949: && (important || time(NULL) - last_io_out >= 5))
950: io_flush(NORMAL_FLUSH);
951: }
952:
953: void maybe_send_keepalive(void)
954: {
955: if (time(NULL) - last_io_out >= allowed_lull) {
956: if (!iobuf_out || !iobuf_out_cnt) {
957: if (protocol_version < 29)
958: send_msg(MSG_DATA, "", 0, 0);
959: else if (protocol_version >= 30)
960: send_msg(MSG_NOOP, "", 0, 0);
961: else {
962: write_int(sock_f_out, cur_flist->used);
963: write_shortint(sock_f_out, ITEM_IS_NEW);
964: }
965: }
966: if (iobuf_out)
967: io_flush(NORMAL_FLUSH);
968: }
969: }
970:
971: void start_flist_forward(int f_in)
972: {
973: assert(iobuf_out != NULL);
974: assert(iobuf_f_out == msg_fd_out);
975: flist_forward_from = f_in;
976: defer_forwarding_messages++;
977: }
978:
979: void stop_flist_forward(void)
980: {
981: flist_forward_from = -1;
982: defer_forwarding_messages--;
983: io_flush(FULL_FLUSH);
984: }
985:
986: /**
987: * Continue trying to read len bytes - don't return until len has been
988: * read.
989: **/
990: static void read_loop(int fd, char *buf, size_t len)
991: {
992: while (len) {
993: int n = read_timeout(fd, buf, len);
994:
995: buf += n;
996: len -= n;
997: }
998: }
999:
1000: /**
1001: * Read from the file descriptor handling multiplexing - return number
1002: * of bytes read.
1003: *
1004: * Never returns <= 0.
1005: */
1006: static int readfd_unbuffered(int fd, char *buf, size_t len)
1007: {
1008: size_t msg_bytes;
1009: int tag, cnt = 0;
1010: char line[BIGPATHBUFLEN];
1011:
1012: if (!iobuf_in || fd != iobuf_f_in)
1013: return read_timeout(fd, buf, len);
1014:
1015: if (!io_multiplexing_in && iobuf_in_remaining == 0) {
1016: iobuf_in_remaining = read_timeout(fd, iobuf_in, iobuf_in_siz);
1017: iobuf_in_ndx = 0;
1018: }
1019:
1020: while (cnt == 0) {
1021: if (iobuf_in_remaining) {
1022: len = MIN(len, iobuf_in_remaining);
1023: memcpy(buf, iobuf_in + iobuf_in_ndx, len);
1024: iobuf_in_ndx += len;
1025: iobuf_in_remaining -= len;
1026: cnt = len;
1027: break;
1028: }
1029:
1030: read_loop(fd, line, 4);
1031: tag = IVAL(line, 0);
1032:
1033: msg_bytes = tag & 0xFFFFFF;
1034: tag = (tag >> 24) - MPLEX_BASE;
1035:
1036: switch (tag) {
1037: case MSG_DATA:
1038: if (msg_bytes > iobuf_in_siz) {
1039: if (!(iobuf_in = realloc_array(iobuf_in, char,
1040: msg_bytes)))
1041: out_of_memory("readfd_unbuffered");
1042: iobuf_in_siz = msg_bytes;
1043: }
1044: read_loop(fd, iobuf_in, msg_bytes);
1045: iobuf_in_remaining = msg_bytes;
1046: iobuf_in_ndx = 0;
1047: break;
1048: case MSG_NOOP:
1049: if (msg_bytes != 0)
1050: goto invalid_msg;
1051: if (am_sender)
1052: maybe_send_keepalive();
1053: break;
1054: case MSG_IO_ERROR:
1055: if (msg_bytes != 4)
1056: goto invalid_msg;
1057: read_loop(fd, line, msg_bytes);
1058: send_msg_int(MSG_IO_ERROR, IVAL(line, 0));
1059: io_error |= IVAL(line, 0);
1060: break;
1061: case MSG_DELETED:
1062: if (msg_bytes >= sizeof line)
1063: goto overflow;
1064: #ifdef ICONV_OPTION
1065: if (ic_recv != (iconv_t)-1) {
1066: xbuf outbuf, inbuf;
1067: char ibuf[512];
1068: int add_null = 0;
1069:
1070: INIT_CONST_XBUF(outbuf, line);
1071: INIT_XBUF(inbuf, ibuf, 0, -1);
1072:
1073: while (msg_bytes) {
1074: inbuf.len = msg_bytes > sizeof ibuf
1075: ? sizeof ibuf : msg_bytes;
1076: read_loop(fd, inbuf.buf, inbuf.len);
1077: if (!(msg_bytes -= inbuf.len)
1078: && !ibuf[inbuf.len-1])
1079: inbuf.len--, add_null = 1;
1080: if (iconvbufs(ic_send, &inbuf, &outbuf,
1081: ICB_INCLUDE_BAD | ICB_INCLUDE_INCOMPLETE) < 0)
1082: goto overflow;
1083: }
1084: if (add_null) {
1085: if (outbuf.len == outbuf.size)
1086: goto overflow;
1087: outbuf.buf[outbuf.len++] = '\0';
1088: }
1089: msg_bytes = outbuf.len;
1090: } else
1091: #endif
1092: read_loop(fd, line, msg_bytes);
1093: /* A directory name was sent with the trailing null */
1094: if (msg_bytes > 0 && !line[msg_bytes-1])
1095: log_delete(line, S_IFDIR);
1096: else {
1097: line[msg_bytes] = '\0';
1098: log_delete(line, S_IFREG);
1099: }
1100: break;
1101: case MSG_SUCCESS:
1102: if (msg_bytes != 4) {
1103: invalid_msg:
1104: rprintf(FERROR, "invalid multi-message %d:%ld [%s]\n",
1105: tag, (long)msg_bytes, who_am_i());
1106: exit_cleanup(RERR_STREAMIO);
1107: }
1108: read_loop(fd, line, msg_bytes);
1109: successful_send(IVAL(line, 0));
1110: break;
1111: case MSG_NO_SEND:
1112: if (msg_bytes != 4)
1113: goto invalid_msg;
1114: read_loop(fd, line, msg_bytes);
1115: send_msg_int(MSG_NO_SEND, IVAL(line, 0));
1116: break;
1117: case MSG_INFO:
1118: case MSG_ERROR:
1119: case MSG_ERROR_XFER:
1120: case MSG_WARNING:
1121: if (msg_bytes >= sizeof line) {
1122: overflow:
1123: rprintf(FERROR,
1124: "multiplexing overflow %d:%ld [%s]\n",
1125: tag, (long)msg_bytes, who_am_i());
1126: exit_cleanup(RERR_STREAMIO);
1127: }
1128: read_loop(fd, line, msg_bytes);
1129: rwrite((enum logcode)tag, line, msg_bytes, 1);
1130: if (first_message) {
1131: if (list_only && !am_sender && tag == 1) {
1132: line[msg_bytes] = '\0';
1133: check_for_d_option_error(line);
1134: }
1135: first_message = 0;
1136: }
1137: break;
1138: default:
1139: rprintf(FERROR, "unexpected tag %d [%s]\n",
1140: tag, who_am_i());
1141: exit_cleanup(RERR_STREAMIO);
1142: }
1143: }
1144:
1145: if (iobuf_in_remaining == 0)
1146: io_flush(NORMAL_FLUSH);
1147:
1148: return cnt;
1149: }
1150:
1151: /* Do a buffered read from fd. Don't return until all N bytes have
1152: * been read. If all N can't be read then exit with an error. */
1153: static void readfd(int fd, char *buffer, size_t N)
1154: {
1155: int cnt;
1156: size_t total = 0;
1157:
1158: while (total < N) {
1159: cnt = readfd_unbuffered(fd, buffer + total, N-total);
1160: total += cnt;
1161: }
1162:
1163: if (fd == write_batch_monitor_in) {
1164: if ((size_t)write(batch_fd, buffer, total) != total)
1165: exit_cleanup(RERR_FILEIO);
1166: }
1167:
1168: if (fd == flist_forward_from)
1169: writefd(iobuf_f_out, buffer, total);
1170:
1171: if (fd == sock_f_in)
1172: stats.total_read += total;
1173: }
1174:
1175: unsigned short read_shortint(int f)
1176: {
1177: char b[2];
1178: readfd(f, b, 2);
1179: return (UVAL(b, 1) << 8) + UVAL(b, 0);
1180: }
1181:
1182: int32 read_int(int f)
1183: {
1184: char b[4];
1185: int32 num;
1186:
1187: readfd(f, b, 4);
1188: num = IVAL(b, 0);
1189: #if SIZEOF_INT32 > 4
1190: if (num & (int32)0x80000000)
1191: num |= ~(int32)0xffffffff;
1192: #endif
1193: return num;
1194: }
1195:
1196: int32 read_varint(int f)
1197: {
1198: union {
1199: char b[5];
1200: int32 x;
1201: } u;
1202: uchar ch;
1203: int extra;
1204:
1205: u.x = 0;
1206: readfd(f, (char*)&ch, 1);
1207: extra = int_byte_extra[ch / 4];
1208: if (extra) {
1209: uchar bit = ((uchar)1<<(8-extra));
1210: if (extra >= (int)sizeof u.b) {
1211: rprintf(FERROR, "Overflow in read_varint()\n");
1212: exit_cleanup(RERR_STREAMIO);
1213: }
1214: readfd(f, u.b, extra);
1215: u.b[extra] = ch & (bit-1);
1216: } else
1217: u.b[0] = ch;
1218: #if CAREFUL_ALIGNMENT
1219: u.x = IVAL(u.b,0);
1220: #endif
1221: #if SIZEOF_INT32 > 4
1222: if (u.x & (int32)0x80000000)
1223: u.x |= ~(int32)0xffffffff;
1224: #endif
1225: return u.x;
1226: }
1227:
1228: int64 read_varlong(int f, uchar min_bytes)
1229: {
1230: union {
1231: char b[9];
1232: int64 x;
1233: } u;
1234: char b2[8];
1235: int extra;
1236:
1237: #if SIZEOF_INT64 < 8
1238: memset(u.b, 0, 8);
1239: #else
1240: u.x = 0;
1241: #endif
1242: readfd(f, b2, min_bytes);
1243: memcpy(u.b, b2+1, min_bytes-1);
1244: extra = int_byte_extra[CVAL(b2, 0) / 4];
1245: if (extra) {
1246: uchar bit = ((uchar)1<<(8-extra));
1247: if (min_bytes + extra > (int)sizeof u.b) {
1248: rprintf(FERROR, "Overflow in read_varlong()\n");
1249: exit_cleanup(RERR_STREAMIO);
1250: }
1251: readfd(f, u.b + min_bytes - 1, extra);
1252: u.b[min_bytes + extra - 1] = CVAL(b2, 0) & (bit-1);
1253: #if SIZEOF_INT64 < 8
1254: if (min_bytes + extra > 5 || u.b[4] || CVAL(u.b,3) & 0x80) {
1255: rprintf(FERROR, "Integer overflow: attempted 64-bit offset\n");
1256: exit_cleanup(RERR_UNSUPPORTED);
1257: }
1258: #endif
1259: } else
1260: u.b[min_bytes + extra - 1] = CVAL(b2, 0);
1261: #if SIZEOF_INT64 < 8
1262: u.x = IVAL(u.b,0);
1263: #elif CAREFUL_ALIGNMENT
1264: u.x = IVAL(u.b,0) | (((int64)IVAL(u.b,4))<<32);
1265: #endif
1266: return u.x;
1267: }
1268:
1269: int64 read_longint(int f)
1270: {
1271: #if SIZEOF_INT64 >= 8
1272: char b[9];
1273: #endif
1274: int32 num = read_int(f);
1275:
1276: if (num != (int32)0xffffffff)
1277: return num;
1278:
1279: #if SIZEOF_INT64 < 8
1280: rprintf(FERROR, "Integer overflow: attempted 64-bit offset\n");
1281: exit_cleanup(RERR_UNSUPPORTED);
1282: #else
1283: readfd(f, b, 8);
1284: return IVAL(b,0) | (((int64)IVAL(b,4))<<32);
1285: #endif
1286: }
1287:
1288: void read_buf(int f, char *buf, size_t len)
1289: {
1290: readfd(f,buf,len);
1291: }
1292:
1293: void read_sbuf(int f, char *buf, size_t len)
1294: {
1295: readfd(f, buf, len);
1296: buf[len] = '\0';
1297: }
1298:
1299: uchar read_byte(int f)
1300: {
1301: uchar c;
1302: readfd(f, (char *)&c, 1);
1303: return c;
1304: }
1305:
1306: int read_vstring(int f, char *buf, int bufsize)
1307: {
1308: int len = read_byte(f);
1309:
1310: if (len & 0x80)
1311: len = (len & ~0x80) * 0x100 + read_byte(f);
1312:
1313: if (len >= bufsize) {
1314: rprintf(FERROR, "over-long vstring received (%d > %d)\n",
1315: len, bufsize - 1);
1316: return -1;
1317: }
1318:
1319: if (len)
1320: readfd(f, buf, len);
1321: buf[len] = '\0';
1322: return len;
1323: }
1324:
1325: /* Populate a sum_struct with values from the socket. This is
1326: * called by both the sender and the receiver. */
1327: void read_sum_head(int f, struct sum_struct *sum)
1328: {
1329: int32 max_blength = protocol_version < 30 ? OLD_MAX_BLOCK_SIZE : MAX_BLOCK_SIZE;
1330: sum->count = read_int(f);
1331: if (sum->count < 0) {
1332: rprintf(FERROR, "Invalid checksum count %ld [%s]\n",
1333: (long)sum->count, who_am_i());
1334: exit_cleanup(RERR_PROTOCOL);
1335: }
1336: sum->blength = read_int(f);
1337: if (sum->blength < 0 || sum->blength > max_blength) {
1338: rprintf(FERROR, "Invalid block length %ld [%s]\n",
1339: (long)sum->blength, who_am_i());
1340: exit_cleanup(RERR_PROTOCOL);
1341: }
1342: sum->s2length = protocol_version < 27 ? csum_length : (int)read_int(f);
1343: if (sum->s2length < 0 || sum->s2length > MAX_DIGEST_LEN) {
1344: rprintf(FERROR, "Invalid checksum length %d [%s]\n",
1345: sum->s2length, who_am_i());
1346: exit_cleanup(RERR_PROTOCOL);
1347: }
1348: sum->remainder = read_int(f);
1349: if (sum->remainder < 0 || sum->remainder > sum->blength) {
1350: rprintf(FERROR, "Invalid remainder length %ld [%s]\n",
1351: (long)sum->remainder, who_am_i());
1352: exit_cleanup(RERR_PROTOCOL);
1353: }
1354: }
1355:
1356: /* Send the values from a sum_struct over the socket. Set sum to
1357: * NULL if there are no checksums to send. This is called by both
1358: * the generator and the sender. */
1359: void write_sum_head(int f, struct sum_struct *sum)
1360: {
1361: static struct sum_struct null_sum;
1362:
1363: if (sum == NULL)
1364: sum = &null_sum;
1365:
1366: write_int(f, sum->count);
1367: write_int(f, sum->blength);
1368: if (protocol_version >= 27)
1369: write_int(f, sum->s2length);
1370: write_int(f, sum->remainder);
1371: }
1372:
1373: /**
1374: * Sleep after writing to limit I/O bandwidth usage.
1375: *
1376: * @todo Rather than sleeping after each write, it might be better to
1377: * use some kind of averaging. The current algorithm seems to always
1378: * use a bit less bandwidth than specified, because it doesn't make up
1379: * for slow periods. But arguably this is a feature. In addition, we
1380: * ought to take the time used to write the data into account.
1381: *
1382: * During some phases of big transfers (file FOO is uptodate) this is
1383: * called with a small bytes_written every time. As the kernel has to
1384: * round small waits up to guarantee that we actually wait at least the
1385: * requested number of microseconds, this can become grossly inaccurate.
1386: * We therefore keep track of the bytes we've written over time and only
1387: * sleep when the accumulated delay is at least 1 tenth of a second.
1388: **/
1389: static void sleep_for_bwlimit(int bytes_written)
1390: {
1391: static struct timeval prior_tv;
1392: static long total_written = 0;
1393: struct timeval tv, start_tv;
1394: long elapsed_usec, sleep_usec;
1395:
1396: #define ONE_SEC 1000000L /* # of microseconds in a second */
1397:
1398: if (!bwlimit_writemax)
1399: return;
1400:
1401: total_written += bytes_written;
1402:
1403: gettimeofday(&start_tv, NULL);
1404: if (prior_tv.tv_sec) {
1405: elapsed_usec = (start_tv.tv_sec - prior_tv.tv_sec) * ONE_SEC
1406: + (start_tv.tv_usec - prior_tv.tv_usec);
1407: total_written -= (int64)elapsed_usec * bwlimit / (ONE_SEC/1024);
1408: if (total_written < 0)
1409: total_written = 0;
1410: }
1411:
1412: sleep_usec = total_written * (ONE_SEC/1024) / bwlimit;
1413: if (sleep_usec < ONE_SEC / 10) {
1414: prior_tv = start_tv;
1415: return;
1416: }
1417:
1418: tv.tv_sec = sleep_usec / ONE_SEC;
1419: tv.tv_usec = sleep_usec % ONE_SEC;
1420: select(0, NULL, NULL, NULL, &tv);
1421:
1422: gettimeofday(&prior_tv, NULL);
1423: elapsed_usec = (prior_tv.tv_sec - start_tv.tv_sec) * ONE_SEC
1424: + (prior_tv.tv_usec - start_tv.tv_usec);
1425: total_written = (sleep_usec - elapsed_usec) * bwlimit / (ONE_SEC/1024);
1426: }
1427:
1428: static const char *what_fd_is(int fd)
1429: {
1430: static char buf[20];
1431:
1432: if (fd == sock_f_out)
1433: return "socket";
1434: else if (fd == msg_fd_out)
1435: return "message fd";
1436: else if (fd == batch_fd)
1437: return "batch file";
1438: else {
1439: snprintf(buf, sizeof buf, "fd %d", fd);
1440: return buf;
1441: }
1442: }
1443:
1444: /* Write len bytes to the file descriptor fd, looping as necessary to get
1445: * the job done and also (in certain circumstances) reading any data on
1446: * msg_fd_in to avoid deadlock.
1447: *
1448: * This function underlies the multiplexing system. The body of the
1449: * application never calls this function directly. */
1450: static void writefd_unbuffered(int fd, const char *buf, size_t len)
1451: {
1452: size_t n, total = 0;
1453: fd_set w_fds, r_fds, e_fds;
1454: int maxfd, count, cnt, using_r_fds;
1455: int defer_inc = 0;
1456: struct timeval tv;
1457:
1458: if (no_flush++)
1459: defer_forwarding_messages++, defer_inc++;
1460:
1461: while (total < len) {
1462: FD_ZERO(&w_fds);
1463: FD_SET(fd, &w_fds);
1464: FD_ZERO(&e_fds);
1465: FD_SET(fd, &e_fds);
1466: maxfd = fd;
1467:
1468: if (msg_fd_in >= 0) {
1469: FD_ZERO(&r_fds);
1470: FD_SET(msg_fd_in, &r_fds);
1471: if (msg_fd_in > maxfd)
1472: maxfd = msg_fd_in;
1473: using_r_fds = 1;
1474: } else
1475: using_r_fds = 0;
1476:
1477: tv.tv_sec = select_timeout;
1478: tv.tv_usec = 0;
1479:
1480: errno = 0;
1481: count = select(maxfd + 1, using_r_fds ? &r_fds : NULL,
1482: &w_fds, &e_fds, &tv);
1483:
1484: if (count <= 0) {
1485: if (count < 0 && errno == EBADF)
1486: exit_cleanup(RERR_SOCKETIO);
1487: check_timeout();
1488: continue;
1489: }
1490:
1491: /*if (FD_ISSET(fd, &e_fds))
1492: rprintf(FINFO, "select exception on fd %d\n", fd); */
1493:
1494: if (using_r_fds && FD_ISSET(msg_fd_in, &r_fds))
1495: read_msg_fd();
1496:
1497: if (!FD_ISSET(fd, &w_fds))
1498: continue;
1499:
1500: n = len - total;
1501: if (bwlimit_writemax && n > bwlimit_writemax)
1502: n = bwlimit_writemax;
1503: cnt = write(fd, buf + total, n);
1504:
1505: if (cnt <= 0) {
1506: if (cnt < 0) {
1507: if (errno == EINTR)
1508: continue;
1509: if (errno == EWOULDBLOCK || errno == EAGAIN) {
1510: msleep(1);
1511: continue;
1512: }
1513: }
1514:
1515: /* Don't try to write errors back across the stream. */
1516: if (fd == sock_f_out)
1517: io_end_multiplex_out();
1518: /* Don't try to write errors down a failing msg pipe. */
1519: if (am_server && fd == msg_fd_out)
1520: exit_cleanup(RERR_STREAMIO);
1521: rsyserr(FERROR, errno,
1522: "writefd_unbuffered failed to write %ld bytes to %s [%s]",
1523: (long)len, what_fd_is(fd), who_am_i());
1524: /* If the other side is sending us error messages, try
1525: * to grab any messages they sent before they died. */
1526: while (!am_server && fd == sock_f_out && io_multiplexing_in) {
1527: char buf[1024];
1528: set_io_timeout(30);
1529: ignore_timeout = 0;
1530: readfd_unbuffered(sock_f_in, buf, sizeof buf);
1531: }
1532: exit_cleanup(RERR_STREAMIO);
1533: }
1534:
1535: total += cnt;
1536: defer_forwarding_messages++, defer_inc++;
1537:
1538: if (fd == sock_f_out) {
1539: if (io_timeout || am_generator)
1540: last_io_out = time(NULL);
1541: sleep_for_bwlimit(cnt);
1542: }
1543: }
1544:
1545: no_flush--;
1546: if (keep_defer_forwarding)
1547: defer_inc--;
1548: if (!(defer_forwarding_messages -= defer_inc) && !no_flush)
1549: msg_flush();
1550: }
1551:
1552: int io_flush(int flush_it_all)
1553: {
1554: int flushed_something = 0;
1555:
1556: if (no_flush)
1557: return 0;
1558:
1559: if (iobuf_out_cnt) {
1560: if (io_multiplexing_out)
1561: mplex_write(sock_f_out, MSG_DATA, iobuf_out, iobuf_out_cnt, 0);
1562: else
1563: writefd_unbuffered(iobuf_f_out, iobuf_out, iobuf_out_cnt);
1564: iobuf_out_cnt = 0;
1565: flushed_something = 1;
1566: }
1567:
1568: if (flush_it_all && !defer_forwarding_messages && msg_queue.head) {
1569: msg_flush();
1570: flushed_something = 1;
1571: }
1572:
1573: return flushed_something;
1574: }
1575:
1576: static void writefd(int fd, const char *buf, size_t len)
1577: {
1578: if (fd == sock_f_out)
1579: stats.total_written += len;
1580:
1581: if (fd == write_batch_monitor_out)
1582: writefd_unbuffered(batch_fd, buf, len);
1583:
1584: if (!iobuf_out || fd != iobuf_f_out) {
1585: writefd_unbuffered(fd, buf, len);
1586: return;
1587: }
1588:
1589: while (len) {
1590: int n = MIN((int)len, IO_BUFFER_SIZE - iobuf_out_cnt);
1591: if (n > 0) {
1592: memcpy(iobuf_out+iobuf_out_cnt, buf, n);
1593: buf += n;
1594: len -= n;
1595: iobuf_out_cnt += n;
1596: }
1597:
1598: if (iobuf_out_cnt == IO_BUFFER_SIZE)
1599: io_flush(NORMAL_FLUSH);
1600: }
1601: }
1602:
1603: void write_shortint(int f, unsigned short x)
1604: {
1605: char b[2];
1606: b[0] = (char)x;
1607: b[1] = (char)(x >> 8);
1608: writefd(f, b, 2);
1609: }
1610:
1611: void write_int(int f, int32 x)
1612: {
1613: char b[4];
1614: SIVAL(b, 0, x);
1615: writefd(f, b, 4);
1616: }
1617:
1618: void write_varint(int f, int32 x)
1619: {
1620: char b[5];
1621: uchar bit;
1622: int cnt = 4;
1623:
1624: SIVAL(b, 1, x);
1625:
1626: while (cnt > 1 && b[cnt] == 0)
1627: cnt--;
1628: bit = ((uchar)1<<(7-cnt+1));
1629: if (CVAL(b, cnt) >= bit) {
1630: cnt++;
1631: *b = ~(bit-1);
1632: } else if (cnt > 1)
1633: *b = b[cnt] | ~(bit*2-1);
1634: else
1635: *b = b[cnt];
1636:
1637: writefd(f, b, cnt);
1638: }
1639:
1640: void write_varlong(int f, int64 x, uchar min_bytes)
1641: {
1642: char b[9];
1643: uchar bit;
1644: int cnt = 8;
1645:
1646: SIVAL(b, 1, x);
1647: #if SIZEOF_INT64 >= 8
1648: SIVAL(b, 5, x >> 32);
1649: #else
1650: if (x <= 0x7FFFFFFF && x >= 0)
1651: memset(b + 5, 0, 4);
1652: else {
1653: rprintf(FERROR, "Integer overflow: attempted 64-bit offset\n");
1654: exit_cleanup(RERR_UNSUPPORTED);
1655: }
1656: #endif
1657:
1658: while (cnt > min_bytes && b[cnt] == 0)
1659: cnt--;
1660: bit = ((uchar)1<<(7-cnt+min_bytes));
1661: if (CVAL(b, cnt) >= bit) {
1662: cnt++;
1663: *b = ~(bit-1);
1664: } else if (cnt > min_bytes)
1665: *b = b[cnt] | ~(bit*2-1);
1666: else
1667: *b = b[cnt];
1668:
1669: writefd(f, b, cnt);
1670: }
1671:
1672: /*
1673: * Note: int64 may actually be a 32-bit type if ./configure couldn't find any
1674: * 64-bit types on this platform.
1675: */
1676: void write_longint(int f, int64 x)
1677: {
1678: char b[12], * const s = b+4;
1679:
1680: SIVAL(s, 0, x);
1681: if (x <= 0x7FFFFFFF && x >= 0) {
1682: writefd(f, s, 4);
1683: return;
1684: }
1685:
1686: #if SIZEOF_INT64 < 8
1687: rprintf(FERROR, "Integer overflow: attempted 64-bit offset\n");
1688: exit_cleanup(RERR_UNSUPPORTED);
1689: #else
1690: memset(b, 0xFF, 4);
1691: SIVAL(s, 4, x >> 32);
1692: writefd(f, b, 12);
1693: #endif
1694: }
1695:
1696: void write_buf(int f, const char *buf, size_t len)
1697: {
1698: writefd(f,buf,len);
1699: }
1700:
1701: /** Write a string to the connection */
1702: void write_sbuf(int f, const char *buf)
1703: {
1704: writefd(f, buf, strlen(buf));
1705: }
1706:
1707: void write_byte(int f, uchar c)
1708: {
1709: writefd(f, (char *)&c, 1);
1710: }
1711:
1712: void write_vstring(int f, const char *str, int len)
1713: {
1714: uchar lenbuf[3], *lb = lenbuf;
1715:
1716: if (len > 0x7F) {
1717: if (len > 0x7FFF) {
1718: rprintf(FERROR,
1719: "attempting to send over-long vstring (%d > %d)\n",
1720: len, 0x7FFF);
1721: exit_cleanup(RERR_PROTOCOL);
1722: }
1723: *lb++ = len / 0x100 + 0x80;
1724: }
1725: *lb = len;
1726:
1727: writefd(f, (char*)lenbuf, lb - lenbuf + 1);
1728: if (len)
1729: writefd(f, str, len);
1730: }
1731:
1732: /* Send a file-list index using a byte-reduction method. */
1733: void write_ndx(int f, int32 ndx)
1734: {
1735: static int32 prev_positive = -1, prev_negative = 1;
1736: int32 diff, cnt = 0;
1737: char b[6];
1738:
1739: if (protocol_version < 30 || read_batch) {
1740: write_int(f, ndx);
1741: return;
1742: }
1743:
1744: /* Send NDX_DONE as a single-byte 0 with no side effects. Send
1745: * negative nums as a positive after sending a leading 0xFF. */
1746: if (ndx >= 0) {
1747: diff = ndx - prev_positive;
1748: prev_positive = ndx;
1749: } else if (ndx == NDX_DONE) {
1750: *b = 0;
1751: writefd(f, b, 1);
1752: return;
1753: } else {
1754: b[cnt++] = (char)0xFF;
1755: ndx = -ndx;
1756: diff = ndx - prev_negative;
1757: prev_negative = ndx;
1758: }
1759:
1760: /* A diff of 1 - 253 is sent as a one-byte diff; a diff of 254 - 32767
1761: * or 0 is sent as a 0xFE + a two-byte diff; otherwise we send 0xFE
1762: * & all 4 bytes of the (non-negative) num with the high-bit set. */
1763: if (diff < 0xFE && diff > 0)
1764: b[cnt++] = (char)diff;
1765: else if (diff < 0 || diff > 0x7FFF) {
1766: b[cnt++] = (char)0xFE;
1767: b[cnt++] = (char)((ndx >> 24) | 0x80);
1768: b[cnt++] = (char)ndx;
1769: b[cnt++] = (char)(ndx >> 8);
1770: b[cnt++] = (char)(ndx >> 16);
1771: } else {
1772: b[cnt++] = (char)0xFE;
1773: b[cnt++] = (char)(diff >> 8);
1774: b[cnt++] = (char)diff;
1775: }
1776: writefd(f, b, cnt);
1777: }
1778:
1779: /* Receive a file-list index using a byte-reduction method. */
1780: int32 read_ndx(int f)
1781: {
1782: static int32 prev_positive = -1, prev_negative = 1;
1783: int32 *prev_ptr, num;
1784: char b[4];
1785:
1786: if (protocol_version < 30)
1787: return read_int(f);
1788:
1789: readfd(f, b, 1);
1790: if (CVAL(b, 0) == 0xFF) {
1791: readfd(f, b, 1);
1792: prev_ptr = &prev_negative;
1793: } else if (CVAL(b, 0) == 0)
1794: return NDX_DONE;
1795: else
1796: prev_ptr = &prev_positive;
1797: if (CVAL(b, 0) == 0xFE) {
1798: readfd(f, b, 2);
1799: if (CVAL(b, 0) & 0x80) {
1800: b[3] = CVAL(b, 0) & ~0x80;
1801: b[0] = b[1];
1802: readfd(f, b+1, 2);
1803: num = IVAL(b, 0);
1804: } else
1805: num = (UVAL(b,0)<<8) + UVAL(b,1) + *prev_ptr;
1806: } else
1807: num = UVAL(b, 0) + *prev_ptr;
1808: *prev_ptr = num;
1809: if (prev_ptr == &prev_negative)
1810: num = -num;
1811: return num;
1812: }
1813:
1814: /* Read a line of up to bufsiz-1 characters into buf. Strips
1815: * the (required) trailing newline and all carriage returns.
1816: * Returns 1 for success; 0 for I/O error or truncation. */
1817: int read_line_old(int f, char *buf, size_t bufsiz)
1818: {
1819: bufsiz--; /* leave room for the null */
1820: while (bufsiz > 0) {
1821: buf[0] = 0;
1822: read_buf(f, buf, 1);
1823: if (buf[0] == 0)
1824: return 0;
1825: if (buf[0] == '\n')
1826: break;
1827: if (buf[0] != '\r') {
1828: buf++;
1829: bufsiz--;
1830: }
1831: }
1832: *buf = '\0';
1833: return bufsiz > 0;
1834: }
1835:
1836: void io_printf(int fd, const char *format, ...)
1837: {
1838: va_list ap;
1839: char buf[BIGPATHBUFLEN];
1840: int len;
1841:
1842: va_start(ap, format);
1843: len = vsnprintf(buf, sizeof buf, format, ap);
1844: va_end(ap);
1845:
1846: if (len < 0)
1847: exit_cleanup(RERR_STREAMIO);
1848:
1849: if (len > (int)sizeof buf) {
1850: rprintf(FERROR, "io_printf() was too long for the buffer.\n");
1851: exit_cleanup(RERR_STREAMIO);
1852: }
1853:
1854: write_sbuf(fd, buf);
1855: }
1856:
1857: /** Setup for multiplexing a MSG_* stream with the data stream. */
1858: void io_start_multiplex_out(void)
1859: {
1860: io_flush(NORMAL_FLUSH);
1861: io_start_buffering_out(sock_f_out);
1862: io_multiplexing_out = 1;
1863: }
1864:
1865: /** Setup for multiplexing a MSG_* stream with the data stream. */
1866: void io_start_multiplex_in(void)
1867: {
1868: io_flush(NORMAL_FLUSH);
1869: io_start_buffering_in(sock_f_in);
1870: io_multiplexing_in = 1;
1871: }
1872:
1873: /** Write an message to the multiplexed data stream. */
1874: int io_multiplex_write(enum msgcode code, const char *buf, size_t len, int convert)
1875: {
1876: if (!io_multiplexing_out)
1877: return 0;
1878: io_flush(NORMAL_FLUSH);
1879: stats.total_written += (len+4);
1880: mplex_write(sock_f_out, code, buf, len, convert);
1881: return 1;
1882: }
1883:
1884: void io_end_multiplex_in(void)
1885: {
1886: io_multiplexing_in = 0;
1887: io_end_buffering_in();
1888: }
1889:
1890: /** Stop output multiplexing. */
1891: void io_end_multiplex_out(void)
1892: {
1893: io_multiplexing_out = 0;
1894: io_end_buffering_out();
1895: }
1896:
1897: void start_write_batch(int fd)
1898: {
1899: /* Some communication has already taken place, but we don't
1900: * enable batch writing until here so that we can write a
1901: * canonical record of the communication even though the
1902: * actual communication so far depends on whether a daemon
1903: * is involved. */
1904: write_int(batch_fd, protocol_version);
1905: if (protocol_version >= 30)
1906: write_byte(batch_fd, compat_flags);
1907: write_int(batch_fd, checksum_seed);
1908:
1909: if (am_sender)
1910: write_batch_monitor_out = fd;
1911: else
1912: write_batch_monitor_in = fd;
1913: }
1914:
1915: void stop_write_batch(void)
1916: {
1917: write_batch_monitor_out = -1;
1918: write_batch_monitor_in = -1;
1919: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>