Hi, Have your machines had prelink turned on? If so, turned it off, and run "prelink -auv" to unlink all binaries or libraries first.
From my experience, one out of 10 cfengine will crash with Seg fault when prelink is on. For these clients, a simple "prelink -auv" fixes it. There was another memory-debugging program also felt uneasy with prelink, I forget its name now but you can google it. Unfortunately cfengine seems having the problem as well, at least for cfengine 2.1.3--2.1.11. Give the "prelink -auv" a try first and let us know the result if your prelink is on. Forget my email if you have not installed prelink at all. Thanks. --Guolin -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Dan Gilbert Sent: Thursday, March 03, 2005 5:25 PM To: David Masterson Cc: [email protected] Subject: RE: Seg faulting cfagent in v.2.1.13 Here's what strace has to say just before the seg fault: poll([{fd=4, events=POLLOUT, revents=POLLOUT}], 1, 0) = 1 send(4, "/n\1\0\0\1\0\0\0\0\0\0\10macgyver"..., 42, 0) = 42 poll([{fd=4, events=POLLIN, revents=POLLIN}], 1, 5000) = 1 ioctl(4, FIONREAD, [58]) = 0 recvfrom(4, "/n\205\200\0\1\0\1\0\0\0\0\10macgyver"..., 1024, 0, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("xxx.xx.xx.xx")}, [16]) = 58 close(4) = 0 write(2, "\320\305\376\277\33", 5? = 5 write(2, ": ", 2: ) = 2 write(2, "Successful return: 0", 20Successful return: 0) = 20 write(2, "\n", 1 ) = 1 --- SIGSEGV (Segmentation fault) @ 0 (0) --- +++ killed by SIGSEGV +++ Dan Dan Gilbert, GCIH, MCSE, CCA Sr. Systems Engineer Advanced iTV Systems/Production Operations Digeo, Inc. [EMAIL PROTECTED] -----Original Message----- From: David Masterson [mailto:[EMAIL PROTECTED] Sent: Thursday, March 03, 2005 5:22 PM To: Dan Gilbert Cc: [email protected] Subject: Re: Seg faulting cfagent in v.2.1.13 You could try 'strace' David Masterson Symbol Technologies >>> "Dan Gilbert" <[EMAIL PROTECTED]> 03/03/05 04:58PM >>> Kind folks of the list -- I've been trying to set up a new cfengine installation, now that I finally have time to do so (though I've needed it for months), I am running into a seg fault situation when running cfagent on a client machine. Both the client and master machines are running Red Hat EL3 with the 2.4.21-27.0.2 kernel and the binaries were built with gcc-3.2.3-49 and glibc-2.3.2-95.30. It looks like things run just fine as the process generates a return code of 0, but then cfagent seg faults and nothing seems to have happened. Any thoughts? My cursory scan of the list archives didn't find anything, though I may not have seen everything, either. Below is the gdb trace as prescribed on the cfwiki.org site (pruned to the area just before the seg fault): (gdb) bt #0 LastSeen (hostname=0xbfff1460 "macgyver", role=cf_connect) at ip.c:422 #1 0x0805cb57 in RemoteConnect (host=0xbfff1460 "macgyver", forceipv4=110 'n') at ip.c:192 #2 0x0805a828 in OpenServerConnection (ip=0x94e69c0) at client.c:57 #3 0x08055081 in MakeImages () at do.c:2425 #4 0x0804d8c8 in DoTree (passes=1, info=0x80a55f8 "Update") at cfagent.c:1284 #5 0x0804b32a in main (argc=-1218542592, argv=0xb75e8400) at cfagent.c:109 #6 0x004d679d in ?? () #7 0x00000002 in ?? () #8 0xbfffb614 in ?? () #9 0xbfffb620 in ?? () #10 0x00000000 in ?? () Thanks, Dan Gilbert, GCIH, MCSE, CCA Sr. Systems Engineer Advanced iTV Systems/Production Operations Digeo, Inc. [EMAIL PROTECTED] ________________________________________________________________________ This email has been scanned for computer viruses. _______________________________________________ Bug-cfengine mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-cfengine _______________________________________________ Bug-cfengine mailing list [email protected] http://lists.gnu.org/mailman/listinfo/bug-cfengine
