Re: Rsync error on client end: unexpected tag 3 [sender] rsync error: error in rsync protocol data stream (code 12) at io.c(843) [sender]

[Wayne Davison]

On Wed, May 03, 2006 at 11:58:07AM +1000, Anthony Sadler wrote:
> Notice /etc/localtime not being able to open? 

That happens because you're using "use chroot = yes".  When rsync is
running inside a chroot jail, it cannot access files that exist outside
the rsync module's path.

Rsync attempts to handle the timezone situation by using localtime()
prior to the chroot call, but on some systems this is not enough to keep
the library code really happy.  I've checked in a change that makes
rsync also call strftime() prior to the chroot (for those systems that
are configured to use strftime()).

..wayne..
--- log.c   25 Apr 2006 23:51:15 -  1.146
+++ log.c   3 May 2006 06:06:19 -
@@ -154,17 +154,14 @@ static void logfile_open(void)
 
 void log_init(void)
 {
-   time_t t;
-
if (log_initialised)
return;
log_initialised = 1;
 
/* this looks pointless, but it is needed in order for the
 * C library on some systems to fetch the timezone info
 * before the chroot */
-   t = time(NULL);
-   localtime(&t);
+   timestring(time(NULL));
 
/* optionally use a log file instead of syslog */
logfname = lp_log_file();
-- 
To unsubscribe or change options: https://lists.samba.org/mailman/listinfo/rsync
Before posting, read: http://www.catb.org/~esr/faqs/smart-questions.html

Re: Rsync error on client end: unexpected tag 3 [sender] rsync error: error in rsync protocol data stream (code 12) at io.c(843) [sender]

[Wayne Davison]

On Wed, May 03, 2006 at 11:58:07AM +1000, Anthony Sadler wrote:
> unexpected tag 3 [sender]

My analysis shows that some deferred messages (which are those that
arrive in the generator when the generator is already trying to flush a
partial message down the socket) are not being handled by rwrite() like
they should be.  Furthermore, the MSG_LOG message doesn't actually need
to be deferred, since they cannot result in any data being written to
the socket.  This bug can only affect an rsync daemon that is receiving
files, as all other scenarios behave correctly with the existing code.

The attached patch should fix the problem.  I've also checked the fix
into CVS (and it will appear in the next nightly tar file that gets
generated).

Thanks for the detailed report, it was very helpful.

..wayne..
--- io.c25 Apr 2006 23:51:15 -  1.193
+++ io.c3 May 2006 05:13:31 -
@@ -294,6 +294,7 @@ static void read_msg_fd(void)
exit_cleanup(RERR_STREAMIO);
}
close_multiplexing_out();
+   defer_forwarding_messages = 0;
/* FALL THROUGH */
case MSG_INFO:
case MSG_ERROR:
@@ -303,7 +304,8 @@ static void read_msg_fd(void)
if (n >= sizeof buf)
n = sizeof buf - 1;
read_loop(fd, buf, n);
-   if (am_generator && am_server && 
defer_forwarding_messages)
+   if (am_generator && am_server
+&& defer_forwarding_messages && tag != MSG_LOG)
msg_list_add(&msg2sndr, tag, buf, n);
else
rwrite((enum logcode)tag, buf, n);
@@ -1134,11 +1136,20 @@ static void msg2sndr_flush(void)
 
while (msg2sndr.head && io_multiplexing_out) {
struct msg_list_item *m = msg2sndr.head;
+   int tag = (IVAL(m->buf, 0) >> 24) - MPLEX_BASE;
if (!(msg2sndr.head = m->next))
msg2sndr.tail = NULL;
-   stats.total_written += m->len;
defer_forwarding_messages = 1;
-   writefd_unbuffered(sock_f_out, m->buf, m->len);
+   switch (tag) {
+   case MSG_INFO:
+   case MSG_ERROR:
+   rwrite((enum logcode)tag, m->buf + 4, m->len - 4);
+   break;
+   default:
+   stats.total_written += m->len;
+   writefd_unbuffered(sock_f_out, m->buf, m->len);
+   break;
+   }
defer_forwarding_messages = 0;
free(m);
}
-- 
To unsubscribe or change options: https://lists.samba.org/mailman/listinfo/rsync
Before posting, read: http://www.catb.org/~esr/faqs/smart-questions.html