[syslog-ng]do_recv() problem with 1.4.9

[Balazs Scheidler]

> > I just upgraded from 1.4.8 to 1.4.9. I compiled with the default options and
> > I'm using the same configuration file as I did with 1.4.8. The only thing that
> > has changed is the syslog-ng binary itself. When I started 1.4.9, it wasn't
> > logging normal events. I have the "internal" messages logging to a separate
> > file, so I checked that. In the few minutes that had passed since I started the
> > daemon, the file had filled up to 250 meg with continuous repetitions of this
> > message:
> > 
> > Nov 22 12:43:30 flogm syslog-ng[27074]: io.c: do_recv(): Zero length read was
> > requested.
> 
> it's silly, can you send me your configuration file? I hardly changed
> anything in 1.4.9. And it passed my simple tests.

Never mind. I could successfully reproduce it. I commented out a HP-UX
workaround, which as it seems was not used only under HP-UX, and I thought
it took part in the DoS I fixed in 1.4.9. Here's the patch, and I'll
promptly release 1.4.9a before the bugtraq announcement goes out.

diff -u -r1.23.2.5 sources.c
--- sources.c	2000/11/22 16:04:59	1.23.2.5
+++ sources.c	2000/11/23 08:29:06
@@ -100,15 +100,12 @@
 	eol = memchr(closure->buffer, '\0', closure->pos);
 	if (eol == NULL)
 		eol = memchr(closure->buffer, '\n', closure->pos);
-#if 0
-	/* this is broken, and breaks other platforms */
-	if (!eol && closure->pos) {
-		/* HP-UX kludge workaround */
+	if (!eol && closure->pos == MAX_LINE) {
+		/* our buffer is full, we have to flush it */
 		do_handle_line(closure, closure->pos, closure->buffer, salen ? (abstract_addr *) &sabuf : NULL, salen);
 		closure->pos = 0;
 		return ST_OK | ST_GOON;
 	}
-#endif
 	start = closure->buffer;
 	while (eol) {
 		/* eol points at the newline character. end points at the


-- 
Bazsi
PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1
     url: http://www.balabit.hu/pgpkey.txt