similar to: rsync 2.5.4 (probably 2.5.5 too) server handles SIGPIPE very poorly

Hi, we ran into a little problem with rsync-2.5.5. Setup: you run rsync-2.5.5 as normal rsync over ssh (ie. not connecting to a rsync server). If you start such a rsync but interrupt the pulling process with Ctrl-C, the process on the other side may start to allocate all memory on the remote machine. As fa as we have analyzed the problem, the remote rsync process wants to issue a error message

-----BEGIN PGP SIGNED MESSAGE----- I added the following code to log_exit(): void log_exit(int code, const char *file, int line) { static int error_count=0; if(error_count++ > 10) { abort(); } To force it to bail earlier instead of overflowing the stack. As you can see at frame #50, it is trying to log that the connection went away unexpectantly.

-----BEGIN PGP SIGNED MESSAGE----- With rsync --server, sometimes when the remote side control-C's a transfer, rsync will immediately start consuming all available RAM. gdb shows that its recursing while trying to report the failed write: #0 0x420503ac in vfprintf () from /lib/i686/libc.so.6 #1 0x42072f74 in vsnprintf () from /lib/i686/libc.so.6 #2 0x08052ca4 in rprintf

(problem encountered on Tru64Unix v4.0f with rsync 2.5.5) A failure to write to stdout or stderr in rwrite (log.c:279) will trigger an endless loop as follows: 0 rprintf calls rwrite 1 rwrite calls fwrite 2 fwrite fails 3 rwrite calls _exit_cleanup 4 _exit_cleanup calls log_exit 5 log_exit calls rprintf 6 rprintf calls rwrite 7

While nbdkit itself must run with SIGPIPE ignored, many applications expect to inherit SIGPIPE in the default state. What's worse, POSIX states that a non-interactive shell script cannot use 'trap' to undo an inherited SIG_IGN on SIGPIPE. I have seen several bug reports over the years of something that works for a developer but fails under a CI environment, where the root cause was

seems the behavior of rsync has changed when dealing with output and using both -v and -q at the same time ... for example: $ mkdir test1 $ touch test1/foo $ rsync-2.6.0 -avq test1 test2 $ rm -r test2 $ rsync-2.6.8 -avq test1 test2 test1/ test1/fo $ rm -r test2 $ rsync-cvs -avq test1 test2 building file list ... test1/ test1/fo $ rm -r test2 the new output in 2.6.8 comes from the calls to

Rsync 2.5.5; transfer via ssh; sparc-sun-solaris2.5 (Ultra 1); gcc 2.8.1. Every time rsync on this box causes boundary violation. I attached two strange examples. I have a few rsync's corefiles (100MB and above) and I can examine them. I will be grateful for any help. -Mirek <-------------------------------------------------------------------------> received 966761 names done

Hello, I have several systems which I recently updated with yum -y update to all the latest packages. These systems use yum-priorities and use the CentOS (priority 1) EPEL (priority 5) and rpmforge (priority 10) repositories. After the updates, dhcpd stopped working with a SIGPIPE error which occurs shortly after it attempts to fork into the background. I worked around that problem by building

Hello, I've got problems with Samba 3.6.0 up to 3.6.6. Forked smbd processes are closing unexpectedly and sometimes smbd server exits. (In Samba 3.5.16 everything works). When invoke smbd from gdb and turn off SIGPIPE passing (by command "handle SIGPIPE nostop nopass") everything works fine. I found out that SigBlk flag of forked smbd processes in /proc/*/status is:

Hi. I'm writing an article on network backups, and instead of using my old ssh1 software, I decided to go with openssh all the way. I got the hang of the openssh way of doing protocol 2 public key authentication, but ssh is failing to terminate when a pipe is broken. I am ssh-ing to a remote host and doing a cat or zcat of a dump file, then on the localhost, I'm using restore to extract

Dear OpenSSH gurus! ;-) I recently upgraded from "OpenSSH 2.1.1p4" to "OpenSSH 2.3.0p1" on my Linux 2.2.17 box with OpenSSL 0.9.5a (RedHat 7.0). According to the "ChangeLog", there was a change in SIGPIPE handling: | 20000930 | [...] | - (djm) Ignore SIGPIPEs from serverloop to child. Fixes crashes with | very short lived X connections. Bug report from

Here's a resend of an old patch that is intended to avoid an infinite recursion (ending in a stack overflow) of the rwrite() function getting an error that calls rwrite(), ad naseum. I've only seen this happen when one of the sides dies due to a program error -- in that case, the connection is closed, and when we try to send an error to the other side and it generates an error, the error

Hi everybody, I use rsync v2.5.5 almost without problems, but it dumps core from time to time. Core was generated by `rsync'. Program terminated with signal 11, Segmentation fault. There is excerpt from gdb's bt: #0 0x2810106f in __sfvwrite () from /usr/lib/libc.so.4 #1 0x280fdfe1 in fprintf () from /usr/lib/libc.so.4 #2 0x281002b6 in vfprintf () from /usr/lib/libc.so.4 #3

Rsync-2.5.2 does not gracefully report connection and transfer errors and always properly return with a non-zero exit code, despite many assurances to the contrary in the code and commit logs. It seems a kludge to handle a special case of lost connections to older servers was FAR too aggressive! With '-vvv' I also print the source of the exit_cleanup() call, and optionally with

Hi, One feature missing from rsync, and requested on this list before, is on-the-fly conversion of filename character encoding. For example, I often need to sync files having Hebrew filenames from a UTF-8 system (Linux) to an ISO8859-8 system (Cygwin on Windows 2000 using the non-Unicode Win32 interface). Other circumstances surely abound. Attached is a patch against rsync 2.6.2 that adds an

In case others find this of value, I wrote a patch to rsync 2.5.6 to give rsync in --daemon mode the ability to run a pre-transfer and post-transfer command. These options handle our need to prepare a server to receive files and to do some processing after receiving files. The options for /etc/rsyncd.conf are pretransfer script = /some/command/to/run posttransfer script =

Hi R-devel, tcltk devel, and sqldf devel, The transcript below shows how loading the tcl/tk library in under R causes subprocesses to ignore SIGPIPE. I am including the developer of the (wonderful) sqldf package since it requires tcltk and you might like to make this dependence optional to the user (at least until this is fixed in tcltk). Am I mistaken in calling this a 'bug'? Any

Not sure if this is a bug, so posting here first. If I run: ?? cnt <- 0L ?? while (TRUE) { ? ? ?? cnt <- cnt + 1L ? ? ?? p <- pipe("echo /dev/stdin > /dev/null", open = "w") ? ? ?? writeLines("foobar", p) ? ? ?? tryCatch(close(p), error = function(e) { print(cnt); stop(e)}) ?? } then once cnt gets to around 650, it fails with: ?? [1] 654 ??

-----BEGIN PGP SIGNED MESSAGE----- The FreeSWAN project uses rsync to keep our FTP repository up-to-date. The FTP server is at xs4all.nl, and we rsync to one of their FreeBSD boxes (xs1.xs4all.nl) over SSH. We have been experiencing core dumps from the remote rsync. Initially this was with the XS4ALL provided rsync in /usr/local/bin/rsync. Since I didn't have access to the source code for

Andreas, How right you are! Still, I find it curious that in the context of the while(TRUE) loop, I am allowed to do this 653 times, with failure on the 654th attempt. Perhaps there is something asynchronous going on? If I eliminate the looping, it does indeed fail (as expected) on the first attempt to close the pipe. Regards Ben On 12/6/19 2:04 AM, Andreas Kersting wrote: > Hi