Date: Mon, 15 Nov 1999 00:05:46 -0800 (PST) From: dlowe@pootpoot.com To: freebsd-gnats-submit@freebsd.org Subject: misc/14895: portmap bug (when run with -v flag) Message-ID: <19991115080546.8E46B14C97@hub.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 14895
>Category: misc
>Synopsis: portmap bug (when run with -v flag)
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Mon Nov 15 00:10:01 PST 1999
>Closed-Date:
>Last-Modified:
>Originator: J. David Lowe
>Release: 3.0-RELEASE
>Organization:
pootpoot
>Environment:
FreeBSD bertha.pootpoot.com 3.0-RELEASE FreeBSD 3.0-RELEASE #3: Sun Oct 31 16:21:08 PST 1999 dlowe@bertha.pootpoot.com:/usr/src/sys/compile/BERTHA i386
>Description:
Under certain circumstances, portmap forks endlessly. This was also
reported in PR misc/12607, but mistakenly listed as a ypbind() bug.
Here's the set of things that cause this problem:
portmap actually fork()s before calling getrpcbynumber() and syslog(),
to log. Thus the main portmap process can answer requests while a
child writes log output.
When portmap is run with the -v flag, it logs every incoming request.
Under certain circumstances, getrpcent(), called several times by
getrpcbynumber(), will call _yp_check(), which in turn may call
ypbind(), which in turn talks to portmap.
Doh!
portmap eventually fills up the process table with logging children.
>How-To-Repeat:
/etc/rc.conf:
nisdomainname="FOO"
nis_server_enable="YES"
portmap_enable="YES"
portmap_flags="-v"
You'll need to go into single user mode in order to get the machine
back, since rc can't seem to finish with a full process table.
>Fix:
This is a workaround patch to pmap_check.c, not very nice because it
removes one level of verbosity from the supposedly more verbose output
of portmap -v. However, it should prevent this from happening, and it
illustrates the problem area. As a simpler workaround, don't run
portmap with the -v flag, or don't run NIS. Neither solution is
perfect. This patch is against "$Id: pmap_check.c,v 1.3 1997/10/09
07:17:11 charnier Exp $"
*** pmap_check.c.original Sun Nov 14 23:36:54 1999
--- pmap_check.c Sun Nov 14 23:42:07 1999
***************
*** 242,248 ****
if (prognum == 0) {
progname = "";
! } else if ((rpc = getrpcbynumber((int) prognum))) {
progname = rpc->r_name;
} else {
sprintf(progname = progbuf, "%lu", prognum);
--- 242,250 ----
if (prognum == 0) {
progname = "";
! /* Ironically, it's not safe to call getrpcbynumber() when
! * verboselog is true, resulting in less verbose logs. */
! } else if ((! verboselog) && ((rpc = getrpcbynumber((int) prognum)))) {
progname = rpc->r_name;
} else {
sprintf(progname = progbuf, "%lu", prognum);
>Release-Note:
>Audit-Trail:
>Unformatted:
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19991115080546.8E46B14C97>