sscalsk@nswc-wo.arpa (02/03/87)
Can anybody offer any explanation for savecores inability to save the core
image? It runs on a 750 with BSD4.2 with 1 RA60 and 1 RA81 drive. The
savecore in the rc.local goes to the RA81:
/etc/savecore /usr01/crash/system >/dev/console
The RA81 has plenty of room. No error messages are generated and savecore
indicates that it was successful, yet no core appears in or along the path.
In looking thru the list of 4.2 bugs I came across a bug for alternate system
images but have not applied the patch yet. Our sources have not localisms so I
must be missing something totally simple-minded. Any ideas folks?
[DISCLAIMER: The preceding was brought to you by me and only me]
Stan Scalsky ARPA: sscalsk@nswc-wo.arpa
Compuserve: 70337,2015
BIX: sscalsky
Source: BBJ841moss@BRL.ARPA (02/03/87)
Stan, It could be that you are getting bit by default "coredumpsize" limits. To fix this from a C shell, type "limit coredumpsize unlimited", or type "limit" to check it. -moss.
sscalsk@nswc-wo.arpa (02/04/87)
> It could be that you are getting bit by default "coredumpsize" limits. > To fix this from a C shell, type "limit coredumpsize unlimited", or type > "limit" to check it. > "Gary S. Moss (SLCBR-VLD-V)" moss@BRL.ARPA Thanks for the simple and elegant suggestion. I have looked at the roots environment and coredumpsize was indeed defaulting to 250 kbytes. I have made the change to unlimited and we will see what happens during the next crash. Thanks to all who responded. [DISCLAIMER: The preceding was brought to you by me and only me] Stan Scalsky Naval Surface Weapons Center - White Oak ARPA: sscalsk@nswc-wo.arpa Compuserve: 70337,2015 BIX: sscalsky Source: BBJ841
jb%cs.brown.edu@RELAY.CS.NET (02/04/87)
I don't have a copy of the 4.2 UDA driver handy but I seem to remember that I wrote uddump smetime after 4.2 was released. That routine is the one which is called upon system restart to copy a memory image to a UDA50. Without this, savecore does not have anything to save. Jim
mike@BRL.ARPA (02/05/87)
Alas, the BSD SAVECORE program has nothing to do with normal "core" files that the kernel writes for you. Thus, altering the coredumplimit on your root shell is unrelated. SAVECORE is a program run by /etc/rc or /etc/rc.local at reboot time, to copy the crash dump from swap space (eg, hp0b) into a normal UNIX file in /dirname/crash (where dirname is the first arg to savecore). The file dirname/minfree indicates how much space the filesystem must have for savecore to start, to keep a sick machine from filling the disk with kernel core dumps. -Mike
chris@mimsy.UUCP (02/05/87)
In article <4195@brl-adm.ARPA> sscalsk@nswc-wo.arpa writes: >Can anybody offer any explanation for savecores inability to save the core >image? What core image? (This is a trick question.) >It runs on a 750 with BSD4.2 with 1 RA60 and 1 RA81 drive. The >savecore in the rc.local goes to the RA81: 4.2BSD does not have a dump routine for RA drives, so savecore can never find a system dump to save. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) UUCP: seismo!mimsy!chris ARPA/CSNet: chris@mimsy.umd.edu
chris@mimsy.UUCP (02/05/87)
>>It could be that you are getting bit by default "coredumpsize" limits. >>To fix this from a C shell, type "limit coredumpsize unlimited", or type >>"limit" to check it. >>"Gary S. Moss (SLCBR-VLD-V)" moss@BRL.ARPA /etc/savecore does not go through the regular core dump code, for what should be obvious reasons if you think about it: The regular dump code depends on the kernel working, and kernel crash dumps are made when the kernel is not working. They are therefore in a special (simple) format; savecore is a user program that converts this to something adb can use. Incidentally, `unlimit <resource>' is a shorthand synonym for `limit <resource> unlimited'. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) UUCP: seismo!mimsy!chris ARPA/CSNet: chris@mimsy.umd.edu