os-linux/1087: Rogue process eating up all the available cpu cycles

[Gilad Gam <gi...@vocaltec.com>]

>Number:         1087
>Category:       os-linux
>Synopsis:       Rogue process eating up all the available cpu cycles
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    apache (Apache HTTP Project)
>State:          open
>Class:          sw-bug
>Submitter-Id:   apache
>Arrival-Date:   Tue Sep  2 13:00:04 1997
>Originator:     gilad@vocaltec.com
>Organization:
apache
>Release:        1.2.4
>Environment:
Redhat Linux 4.0, kernel 2.0.25, no special stuff installed. 
Apache complied with gcc 2.7.2, with modules mod_rewrite and mod_proxy in
addition to all the default modules.

uname -a says:
Linux vtoff1.vocaltec.com 2.0.25 #2 Fri Jan 17 17:56:03 EST 1997 i586
>Description:
After running fine for a while, one of the child processes will go berserk, 
take up all available cpu cycles, writing the same error to the disk over and 
over again:
[Fri Aug 29 22:20:59 1997] select: (listen): Bad address

Using strace on the process gives:
write(17, "[Tue Sep  2 14:10:32 1997] selec"..., 57) = 57
oldselect(16, [15], NULL, NULL, NULL)   = -1 EFAULT (Bad address)
time(NULL)                              = 873223832

The server behaves very sluggish after it starts, though nothing really breaks 
and no coredump is generated.

Sending the root process SIGHUP seems to clean the problem up.
>How-To-Repeat:
No easy way, it appears after the server is running for about
13 hours from the last hangup.
>Fix:
Sorry, nope. I don't have enough time to get into the code
>Audit-Trail:
>Unformatted: