Sorry about the csa_halt bug. It has been identified and fixed.
We know that bug had nothing to do with the ja problem you mentioned.
To get around the csa_halt problem, please just recycle the csa
by doing "/etc/init.d/csa stop; /etc/init.d/csa start".
Thanks,
- jay
Andrew Fant wrote:
--On Friday, October 15, 2004 12:42:54 -0500 Robin Holt <holt@xxxxxxx>
wrote:
Nope, there must be something else going on. We would need to know
a lot more to help you. I guess we would need to start with kernel
version and patches, job and csa userland versions, glibc version,
and compiler version.
It will probably be easier to get a core dump and just issue 'where'.
That will probably be the best information.
Robin:
I do have job and csa set to run at boot-time, the pam.d files have
been modified, jobtest runs successfully, and jstat does return a jid
when I log in. I installed job and csa from sources, not from the rpm.
Oddly, I have just discovered a second error mode:
chicken04 root # /usr/sbin/csaswitch -c halt
/proc/csa ioctl failure, command='csa_halt'
System Error(14): Bad address.
Unable to halt system accounting.
System Error(14): Bad address.
As for system specs, I have reproduced the error with both a vanilla
un-patched 2.4.26 kernel and the patched gentoo-sources 2.4.26 kernel.
Glibc is 2.3.3. gcc is version 3.3.3. I am using job 1.4 and csa
userland 2.2.0. For pagg and job I am using the 2.4.26-4 patches.
Thanks for your help so far. If there is anything else you want to
know, please let me know.
Andy
|