Red Hat Bugzilla – Bug 215378
udev eats 100% of cpu when adding 32+ ttys.
Last modified: 2007-11-30 17:07:27 EST
Description of problem:
udev eats 100% of cpu for 5+ minutes when adding 32+ ttys using
Version-Release number of selected component (if applicable):
Latest "up2date" for everything on ES 4.
Steps to Reproduce:
1. Add a driver to system that registers 32 ttys using the tty_register_device
() call in kernel.
2. With or without a "rule" for those devices in /etc/udev/rules.d, it doesn't
System will literally be bogged down for 5-10 minutes, system hasn't panic'ed,
but is horribly unresponsive for those 5-10 minutes.
An "uptime" will report load averages past 100!!!
A "ps -ef" (when it eventually responses) shows 100's and 100's of:
/bin/sh /sbin/hotplug tty
Very minimal cpu usage.
The most important thing here is that under Fedora Core 4, 5 and 6,
this isn't a problem at all!
It works absolutely perfect uder FC 4, 5 and 6!
The udev version that appears to be in use on RH Enterprise 4 is "039".
This is *very* *very* old compared to FC 4, 5 or 6.
udev is currently up to "100" or so now!
I am guessing that something in "udev" has changed in newer versionf of udev
that makes things work a lot crisper.
Why is udev not being kept up on RH 4?
*** This bug has been marked as a duplicate of 215379 ***