mysqld in an infinite loop

[Christopher Chan]

Hi I have found mysqld in an infinite loop.

strace -p pid would bring

break(0xnumber)    = -1 ENOMEM (Cannot allocate memory) repeatedly.

How-To-Repeat: Generate plenty of mysql connections until the machine 
starts swapping. When the bug is triggered, you will not be able to connect 
to mysqld anymore and if you kill off all other processes, you will find 
mysql happily taking up all CPU that is available to run its infinite loop.

This mysql installation has InnoDB enabled and almost all the connections 
run selects on an indexed column in a InnoDB table.

mysqld could not be killed and nor could we connect to it to shut it down.

A reboot of the machine was the only thing we could do. This only happens 
when swap is touched but does not occur straight away.

It runs on a FreeBSD 4.7 Stable box. Swap was not completely used up.

We need to be able to push this box but if mysql is going to into an 
infinite loop during the times it hits swap it is going to pose a rather 
big problem. Is there anything that I can do to track down where this one is?