Skip to contentSkip to conversation

logger: Segmentation fault when reading from stdin and writing to socket

PATCH Patrick Plagwitz 4 participants 5 messages Apr 6, 2015 – Apr 7, 2015 Util-Linux package

lore
Share & tools
lore
#0 lore raw

Hi,

I have noticed a bug within the development version of the logger utility (misc-utils/logger.c).
When reading messages from stdin (so that logger_stdin is called) and outputting to a TCP/UDP socket (inet_socket) logger SEGFAULTs:

$ echo foo | logger -n localhost
Segmentation fault (core dumped)

This bug happens because, in this combination, the syslog header isn't generated before calling strlen() on it and has probably been introduced somewhere when separating writing the header and writing the message. Calling generate_syslog_header in logger_open also for inet sockets fixes this.
The attached patch does just that.

On an unrelated note, the write_output function leaks memory by not freeing the buf local variable.

Patrick

misc-utils/logger.c +7 -6
index edc9483..fbe8ced 100644
@@ -582,14 +582,15 @@ static void logger_open(struct logger_ctl *ctl)
 		ctl->fd = inet_socket(ctl->server, ctl->port, ctl->socket_type);
 		if (!ctl->syslogfp)
 			ctl->syslogfp = syslog_rfc5424_header;
-		return;
+	} else {
+		if (!ctl->unix_socket)
+			ctl->unix_socket = _PATH_DEVLOG;
+
+		ctl->fd = unix_socket(ctl, ctl->unix_socket, ctl->socket_type);
+		if (!ctl->syslogfp)
+			ctl->syslogfp = syslog_local_header;
 	}
-	if (!ctl->unix_socket)
-		ctl->unix_socket = _PATH_DEVLOG;
 
-	ctl->fd = unix_socket(ctl, ctl->unix_socket, ctl->socket_type);
-	if (!ctl->syslogfp)
-		ctl->syslogfp = syslog_local_header;
 	if (!ctl->tag)
 		ctl->tag = xgetlogin();
 	generate_syslog_header(ctl);
#1 lore raw
Quoted 15 lines · Patrick Plagwitz

I have noticed a bug within the development version of the logger utility (misc-utils/logger.c).
When reading messages from stdin (so that logger_stdin is called) and outputting to a TCP/UDP socket (inet_socket) logger SEGFAULTs:

$ echo foo | logger -n localhost
Segmentation fault (core dumped)

This bug happens because, in this combination, the syslog header isn't generated before calling strlen() on it and has probably been introduced somewhere when separating writing the header and writing the message. Calling generate_syslog_header in logger_open also for inet sockets fixes this.
The attached patch does just that.

Quoted 2 lines

On an unrelated note, the write_output function leaks memory by not freeing the buf local variable.

Hi Patrick and others,

The below appears to fix the issue. And this reminds me logger(1) ought to have 'remote' logging tests writing to netcat(1) server.

--->8----
From: Sami Kerola <[email protected]>
Date: Mon, 6 Apr 2015 22:42:54 +0100
Subject: [PATCH] logger: generate header when reading message from stdin

This change fixes crashing error, that ought not to be simply avoided.

$ echo foo | logger -n localhost
Segmentation fault (core dumped)

If the ctl->hdr is just checked not to be NULL syslog message will not have valid header, so generating such is not optional when reading message from stdin and writing it to remote destination.

---
 misc-utils/logger.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
misc-utils/logger.c +7 -7
index edc9483..753cd7f 100644
@@ -582,14 +582,14 @@ static void logger_open(struct logger_ctl *ctl)
 		ctl->fd = inet_socket(ctl->server, ctl->port, ctl->socket_type);
 		if (!ctl->syslogfp)
 			ctl->syslogfp = syslog_rfc5424_header;
-		return;
-	}
-	if (!ctl->unix_socket)
-		ctl->unix_socket = _PATH_DEVLOG;
+	} else {
+		if (!ctl->unix_socket)
+			ctl->unix_socket = _PATH_DEVLOG;
 
-	ctl->fd = unix_socket(ctl, ctl->unix_socket, ctl->socket_type);
-	if (!ctl->syslogfp)
-		ctl->syslogfp = syslog_local_header;
+		ctl->fd = unix_socket(ctl, ctl->unix_socket, ctl->socket_type);
+		if (!ctl->syslogfp)
+			ctl->syslogfp = syslog_local_header;
+	}
 	if (!ctl->tag)
 		ctl->tag = xgetlogin();
 	generate_syslog_header(ctl);
Signature · 1 line
2.3.5
#2 lore raw
Quoted 4 lines · Sami Kerola
Quoted 1 line · Patrick Plagwitz

The attached patch does just that.

The below appears to fix the issue.

Sami, your patch differs from Patrick's just in a newline, and you only turned his - already great - description into a commit message. Therefore I think setting Patrick as the author would be both correct and just fair.

Please note that seeing the own name in the commit log encourages people to continue contributing in future.

BTW: what about the memleak he reported?

Thanks & have a nice day,
Berny

#3 lore raw
Quoted 6 lines · Bernhard Voelker
Quoted 4 lines · Sami Kerola
Quoted 1 line · Patrick Plagwitz

The attached patch does just that.

The below appears to fix the issue.

Quoted 7 lines

Sami, your patch differs from Patrick's just in a newline, and you only turned his - already great - description into a commit message. Therefore I think setting Patrick as the author would be both correct and just fair.

Please note that seeing the own name in the commit log encourages people to continue contributing in future.

Hi Bernhard,

Oh, I did not even notice there was an attachment. What comes to attributing contributors you are absolutely right. Adjusted change can be found from my remote branch logger-fix.

https://github.com/kerolasa/lelux-utiliteetit/commit/4a8919a4e5b28a47cd61fc8b774a6eaee943b90e

Quoted 1 line

BTW: what about the memleak he reported?

Can be fixed. And since starting to look this I took liberty to check with valgrind
if there are other leaks. One more found from logger_command_line().

https://github.com/kerolasa/lelux-utiliteetit/commit/c3dd2ecd5fcaf30860d5fcfd74edfd70a3dd7603

Signature · 2 lines
Sami Kerola
http://www.iki.fi/kerolasa/
maintainer
#4 lore raw
Quoted 5 lines · Sami Kerola
Quoted 2 lines · Bernhard Voelker
Quoted 1 line · Sami Kerola

On Mon, 6 Apr 2015, Patrick Plagwitz wrote:

https://github.com/kerolasa/lelux-utiliteetit/commit/4a8919a4e5b28a47cd61fc8b774a6eaee943b90e
https://github.com/kerolasa/lelux-utiliteetit/commit/c3dd2ecd5fcaf30860d5fcfd74edfd70a3dd7603

Thanks Patrick and Sami.

...by the way -- it means we don't have stdin logger tests at all!

Karel

Signature · 2 lines
 Karel Zak  <[email protected]>
 http://karelzak.blogspot.com