Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 61573 - vixie-cron segfaulting in fdopen() / malloc()
Summary: vixie-cron segfaulting in fdopen() / malloc()
Status: RESOLVED TEST-REQUEST
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: [OLD] Core system (show other bugs)
Hardware: x86 Linux
: High normal
Assignee: Gentoo's Team for Core System packages
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2004-08-24 16:32 UTC by David
Modified: 2004-09-02 22:49 UTC (History)
1 user (show)

See Also:
Package list:
Runtime testing required: ---


Attachments
A gdb backtrace of the corefile dumped by cron (CronBacktrace.txt,2.62 KB, text/plain)
2004-08-24 16:33 UTC, David
Details
A gdb backtrace of the corefile dumped by cron, when running under efence (CronBacktrace.efence.txt,346 bytes, text/plain)
2004-08-24 16:34 UTC, David
Details
Compiled cron binary and two corefiles (Cron.tar.bz2,213.30 KB, application/x-tbz)
2004-08-24 16:40 UTC, David
Details

Note You need to log in before you can comment on or make changes to this bug.
Description David 2004-08-24 16:32:58 UTC
Using vixie-cron-3.0.1-r5 (currently in ~x86, x86 package exhibits same issues).  After building and running on a kernel with grsecurity enabled, segfaults are logged to syslog when the root/system crontab is run.  This is a periodic, reproducible error every time the root/system crontab is run.  Mail from root/system crontabs is not sent but commands are run.

Example output from grsecurity:

Aug 24 23:45:01 xx kernel: grsec: From xx: signal 11 sent to /usr/sbin/cron[cron:27980] uid/euid:0/0 gid/egid:0/0, parent /usr/sbin/cron[cron:7560] uid/euid:0/0 gid/egid:0/0

Note that grsecurity is only logging.  PaX is not enabled, nor are ACLs.  Without grsecurity in the kernel, the same behaviour is exhibited by cron (a la no sending mail for root/system crontabs).

To debug, cron was built with:

# FEATURES="nostrip" CFLAGS="-g3 -ggdb -fno-omit-frame-pointer" emerge vixie-cron

glibc and pam (as the "pam" USE flag is enabled) were rebuilt using the same FEATURES and CFLAGS above, with the "debug" USE flag added also.

My cron binary and the core files should be available attached.  The output from a gdb backtrace on the produced corefile (from when cron segfaults) should also be attached.


The problem appears to be with something calling fdopen(), which then calls malloc().  To test the malloc() issue, vixie-cron was run under efence using:

# LD_PRELOAD=libefence.so.0.0 /usr/sbin/cron

When running under efence, cron exhibits apparent "normal" behaviour and email from root/system crontabs is sent successfully.  However, a core is still dumped which should also be available attached, along with the backtrace from gdb on this corefile.  Whilst the efence corefile seems to point at an error with PAM, building vixie-cron with USE="-pam" still produces the same issues.


Running cron under strace does not seem to segfault or dump core and does send mail correctly.  However, the following message is observed in syslog, only when running cron under strace:

CRON[14513]: (root) MAIL (mailed 4 bytes of output but got status 0xffffffff )


This problem has existed for a number of versions of vixie-cron on this system but was not noticed at the time.  vixie-cron runs on a number of other systems this author runs with no apparent issues similar to this.

Reproducible: Always
Steps to Reproduce:
1. Run vixie-cron using /etc/init.d/vixie-cron start, script modified with `ulimit -c unlimited` before start-stop-daemon
2. Sit back and wait for it to dump core

Actual Results:  
 

Expected Results:  
 

simba root # emerge info 
Portage 2.0.50-r9 (default-x86-1.4, gcc-3.3.3, glibc-2.3.3.20040420-r1, 
2.4.27-grsec-2.0.1) 
================================================================= 
System uname: 2.4.27-grsec-2.0.1 i686 Pentium III (Coppermine) 
Gentoo Base System version 1.4.16 
ccache version 2.3 [enabled] 
Autoconf: sys-devel/autoconf-2.59-r4 
Automake: sys-devel/automake-1.8.5-r1 
ACCEPT_KEYWORDS="x86" 
AUTOCLEAN="yes" 
CFLAGS="-O3 -march=pentium3 -funroll-loops -pipe" 
CHOST="i686-pc-linux-gnu" 
COMPILER="gcc3" 
CONFIG_PROTECT="/etc /usr/X11R6/lib/X11/xkb /usr/kde/2/share/config /usr/kde/3/share/config /usr/share/config /var/qmail/control" 
CONFIG_PROTECT_MASK="/etc/gconf /etc/terminfo /etc/env.d" 
CXXFLAGS="-O3 -march=pentium3 -funroll-loops -pipe" 
DISTDIR="/usr/portage/distfiles" 
FEATURES="autoaddcvs ccache sandbox" 
GENTOO_MIRRORS="http://www.mirrorservice.org/sites/www.ibiblio.org/gentoo/ 
http://ftp.heanet.ie/pub/gentoo/" 
MAKEOPTS="-j2" 
PKGDIR="/packages" 
PORTAGE_TMPDIR="/var/tmp" 
PORTDIR="/usr/portage" 
PORTDIR_OVERLAY="/usr/local/portage" 
SYNC="rsync://rsync.gentoo.org/gentoo-portage" 
USE="apache apache2 apm arts avi berkdb crypt curl encode exiscan exiscan-acl 
flash foomaticdb gd gdbm gif gpm gtk2 imap imlib innodb java jpeg justify ldap libg++ 
libwww mad mailboxlist maildir mcal memlimit mikmod motif mpeg mysql ncurses nls 
oggvorbis opengl oss pam pdflib perl png python quicktime readline sdl slang spell ssl 
svga truetype x86 xml2 xv zlib"
Comment 1 David 2004-08-24 16:33:48 UTC
Created attachment 38125 [details]
A gdb backtrace of the corefile dumped by cron
Comment 2 David 2004-08-24 16:34:32 UTC
Created attachment 38126 [details]
A gdb backtrace of the corefile dumped by cron, when running under efence
Comment 3 David 2004-08-24 16:40:09 UTC
Created attachment 38127 [details]
Compiled cron binary and two corefiles

The compiled binary version of cron and two corefiles.	The corefiles
correspond to the backtraces, also attached.
Comment 4 SpanKY gentoo-dev 2004-08-24 20:36:55 UTC
what if you upgrade to the latest stable gcc and rebuild vixie-cron
Comment 5 David 2004-08-25 02:18:47 UTC
gcc is gcc-3.3.3-r6, unless I am mistaken this is the latest in x86.
Comment 6 SpanKY gentoo-dev 2004-08-30 16:57:09 UTC
true, i thought `emerge info` would have shown the -r# ...

please try vixie-cron-4.1
Comment 7 David 2004-09-02 22:49:06 UTC
Unfortunately, this is a production server that has had a broken vixie-cron on it for a long while therefore we have now moved to dcron.  If an appropriate maintenance window is found, I will attempt to recompile 4.1 and test it.