[squid-users] Segmentation Fault, dlmalloc

Hi,

I'm running squid 2.5 stable 10 on redhat 8.0 machine and I have been
getting many errors regarding segment violation. The following are
some of the information that might be useful:

Linux Redhat 8.0

# uname -a
Linux proxy7 2.4.18-14 #1 Wed Sep 4 13:35:50 EDT 2002 i686 i686 i386 GNU/Linux

# /etc/squid/sbin/squid -v
Squid Cache: Version 2.5.STABLE10
configure options: --prefix=/etc/squid --sysconfdir=/etc/squid
--localstatedir=/logs --enable-snmp --enable-dlmalloc

Information from gdb:

2005/07/23 16:45:32| httpReadReply: Excess data from "GET
http://entier.ecosm.com/img/?wwobyprr"
2005/07/23 16:46:52| parseHttpRequest: Unsupported method 'GNUTELLA'
2005/07/23 16:46:52| clientReadRequest: FD 751 Invalid Request
2005/07/23 16:47:39| httpReadReply: Excess data from "GET
http://entier.ecosm.com/img/?wwobyprr"
2005/07/23 16:48:52| WARNING! Your cache is running out of filedescriptors
2005/07/23 16:49:37| urlParse: Illegal character in hostname
'inter_1908.w.interia.pl'

Program received signal SIGSEGV, Segmentation fault.
0x080c1227 in free (mem=0xbffef730) at dlmalloc.c:2325
2325 nextsz = chunksize(next);
(gdb) backtrace
#0 0x080c1227 in free (mem=0xbffef730) at dlmalloc.c:2325
#1 0x420d649d in vsyslog () from /lib/i686/libc.so.6
#2 0x420d617f in syslog () from /lib/i686/libc.so.6
#3 0x080b2456 in fatal_common (message=0x8180c20 "xcalloc: Unable to
allocate 1 blocks of 4108 bytes!\n") at tools.c:336
#4 0x080b25a2 in fatal_dump (message=0x8180c20 "xcalloc: Unable to
allocate 1 blocks of 4108 bytes!\n") at tools.c:405
#5 0x080c84d2 in xcalloc (n=1, sz=4108) at util.c:561
#6 0x08094ae9 in memPoolAlloc (pool=0x8219220) at MemPool.c:254
#7 0x080939c4 in memAllocate (type=MEM_MEM_NODE) at mem.c:127
#8 0x080a9756 in stmemAppend (mem=0x35eceb08,
    data=0x8119ad1
"ãÓÇ3°\f;(c)Jg\021\006\v4>óV\027\216\234u\232ØËþá\005Ñ+\203\fÊ\222É|oÎâ(c)\\Æ\001<½9aÁ\236^\231äR0wO

qá%øLá·ÑÒ¹|\031È\231Ln\237/Z>_è9_°\034ì<W¼t¾\020

ý\002FE\224»PxûBqÔ\203eJ\003\tÑ%:åVÛ£\2012½Ë/½\037ü£RÖA\212íæå÷\021>²ü¾`v«±|¶\fÛ?
«/Uâ§ø;¼3n;¼y]\037ãý\032\005íýô\023N­\212bq

\024E\\ªÕ#\005íD7Ç_4àUìÌêS?ú;ü\206%Þ"..., len=44270) at stmem.c:119
#9 0x080aaaea in storeAppend (e=0x35ea8638,
    buf=0x81147c0

"Þï¬õß\223ÙêÒ:W\214\030tê\215\2309\037¢\212(c)\017\214ë\023\234[D\2206\235Òæ¸Hô\232h
S7\fìoØp\211\bF&þF\214\227\2102yÄxü>¹4û

\005\027\rç\206ÃÔ\030³{'y_Ô\"è\023¾ËÖüÓ/XÅ\230ù¨½\231\031æuÄñ\024\024Å\006gËÿYN#*\a¼)¼<\016Z'\233\206½\234\201ȵq%#ig\002\rðÍ

É«\231a¡(\207JÙh²²\auñ¾å#\fíÝíL*uäLw\222·\020
\037O0\tU\235@sgo\222ø[¹³DÈ\023aî\026¢\021\237Ðò\e\022?\177"...,
len=65535) at

store.c:529
#10 0x08083796 in httpReadReply (fd=778, data=0x2400d768) at http.c:704
#11 0x0806bcef in comm_poll (msec=438) at comm_select.c:445
#12 0x08092f39 in main (argc=2, argv=0xbffffb14) at main.c:742
#13 0x420158d4 in __libc_start_main () from /lib/i686/libc.so.6

Based on the information acquired, it seems that dlmalloc is not
working well. Besides re-compiling squid without dl-malloc option,
what else can I do to solve this problem? I would like to hear some
alternative solutions/suggestions from the experts before deciding.

I'm not sure whether this have anything to do with the segmentation
fault problem but according to the gdb output, it seems that my cache
server is running out of file descriptors (WARNING! Your cache is
running out of filedescriptors). I was thinking about modifying the
"nofile" value in /etc/security/limits.conf. Will that suffice?

Thank you.
Received on Sat Jul 23 2005 - 08:25:45 MDT