Red Hat Bugzilla – Bug 245291
bind-utils are SegFaulting
Last modified: 2007-11-30 17:12:08 EST
Description of problem: The bind-utils (host, dig, rndc,....) are segfaulting regularly. Version-Release number of selected component (if applicable): 9.4.0-6 How reproducible: Simple shell script which is querying "rndc status" every second. Actual results: SegFault. (Occassionally.) Expected results: No problems. Additional info: The problem first occured on HA systems (heartbeat) where "rndc status" is called every 2 seconds to check the Bind daemon. After some investigations the problem could be identified in the PIE comiler/linker options. Disabling this option solves the problem.
Interesting. This never occurs to me. I'm going to try reproduce it. It would be nice if you attach core dump when next segfault occurs. You could run something like this #!/bin/bash ulimit -c unlimited while [ 1 -eq 1 ]; do sleep 1; /usr/sbin/rndc status; done Thanks, Adam
Created attachment 157603 [details] Core Dump
Thanks, I'm going to analyze it
Realy looks like glibc issue. It will be gcc problem but looks more like dynamic linker problem for me. Backtrace from uploaded core dump (gdb) bt #0 0x0059c402 in _start () from /lib/ld-linux.so.2 Cannot access memory at address 0xbfa1e9f0 Reassigning to glibc Adam
More likely kernel randomization messing up things. The core file is extremely weird: info regs eax 0xffffffea -22 ecx 0x9ae9f58 162439000 edx 0x9ae9090 162435216 ebx 0x9ae8ee0 162434784 esp 0xbfa1e9f0 0xbfa1e9f0 ebp 0xbfa1eaf8 0xbfa1eaf8 esi 0x9ae6d88 162426248 edi 0xbd3ff4 12402676 eip 0x59c402 0x59c402 <_start> eflags 0x212 [ AF IF ] cs 0x73 115 ss 0x7b 123 ds 0x7b 123 es 0xc040007b -1069547397 fs 0x0 0 gs 0x0 0 but the PT_LOAD segments in core are: LOAD 0x001000 0x00a2d000 0x00000000 0x00000 0x05000 R E 0x1000 LOAD 0x001000 0x00a32000 0x00000000 0x02000 0x02000 RW 0x1000 LOAD 0x003000 0x00a78000 0x00000000 0x00000 0x19000 R E 0x1000 LOAD 0x003000 0x00a91000 0x00000000 0x02000 0x02000 RW 0x1000 LOAD 0x005000 0xbff69000 0x00000000 0x15000 0x15000 RW 0x1000 So, neither $eip nor $esp nor $ebp fall into an mapped region. I certainly can't reproduce this myself on F7 with x86_64 kernel, nor on RHEL5 i686, but occasional PIE crashes have been reported already in the past, e.g. #217614.
Steinhauer Juergen, as it is unreproducible for me could you please provide some system info by: cat /proc/version;a=`rpm --qf '%{name}-%{version}-%{release}.%{arch}\n' -qa|egrep '^(kernel|glibc|bind)'|sort`;echo $a;rpm -V $a
Linux version 2.6.21-1.3228.fc7 (kojibuilder@xenbuilder2.fedora.redhat.com) (gcc version 4.1.2 20070502 (Red Hat 4.1.2-12)) #1 SMP Tue Jun 12 15:37:31 EDT 2007 bind-9.4.1-6.1.fc7.i386 bind-libs-9.4.1-6.1.fc7.i386 bind-utils-9.4.1-6.1.fc7.i386 glibc-2.6-3.i686 glibc-common-2.6-3.i386 kernel-2.6.21-1.3194.fc7.i686 kernel-2.6.21-1.3228.fc7.i686 I've made a VMware installation in which I can also reproduce the problem.
*** This bug has been marked as a duplicate of 246623 ***