Weird Slowness with 2.4.5

[kozumo]

I just compiled 2.4.5 (I'm posting from it now) and everything seems to work fine except there's this weird slowness.

1. The CPU monitor in GKrellM is always at 100%, but nothing is running to make it so. I tested this by running genome@home and genome@home would take so long to process. So something is definitely making the CPU work like crazy. Did I compile the wrong feature or module in the kernel?

In 2.2.18 (my backup kernel) this problem doesn't occur and genome@home flies.

2. Well for one thing Opera is slower in its rendering of pages. I did a Ctrl-Alt F1 to see if there were error messages and I noticed that "DRI" is disabled (I included this feature as a module for my ATI-All-In-Wonder 128). The driver seems to load fine. Also, when I do startx, it takes longer to load X than it does with 2.2.18.

I admit that I don't exactly know what all of the features in the kernel do, but I checked the minimal amount of things I needed and whenever 'Help' told me that if I were unsure to check 'N' or 'Y', I'd do it...unless I need it.

I noticed also that init was trying to load 'NLS 8859' or somethign similar to that and it could not find it. So I'm going to try and modprobe it.

How can I get rid of this weirdness with the CPU always at 100%?

-- possible relevant info --

top says: 03:25:32 up 40 min, 2 users, load average: 1.90, 1.97, 1.78
57 processes: 52 sleeping, 5 running, 0 zombie, 0 stopped
CPU states: 48.3% user, 51.7% system, 0.0% nice, 0.0% idle
Mem: 255792K total, 104532K used, 151260K free, 3572K buffers
Swap: 0K total, 0K used, 0K free, 69796K cached

Aha!

Ok, portsentry and syslogd seem to be hogging the CPU almost 50%/50%...

So it's a Networking module I must've compiled wrong?

Yes indeed, I have killed portsentry and the CPU monitor in GKrellM dropped to 0%. Then I restarted it again and it flew up to 100%.

-- added --
Since I know nobody's going to reply at this very late hour, I might as well put my hardware specs:

Athlon 1Ghz:
I chose Athon/Duron option in the CPU Type section.

SB Live!:
chose the emu10k1 module. No problems.

RTL8139:
had to include '8139too' in /etc/modules, otherwise it wouldn't load and I wouldn't have internet.

Other things that I chose are:
dhcp, everything that had to do with NFS, vfat, mostly left usb features at default, scsi emulation and some I don't remember.

Other than that I have a pretty typical system: 256MB RAM, IDE controller, 20GB HD...

[ 25 June 2001: Message edited by: kozumo ]

[Ig0r]

Try turning off portsentry, and checking in /var/log/ for large logfiles.
If you're getting 50% cpu from a logger, then some daemon is probably having a fit and dumping it all to a logfile.

I don't know about portsentry, but it might require ipchains compiled in your kernel. Check your 2.4.* config to see if you have ipchains enabled (Y/M).

[Xsecrets]

you could always run top and see whats using all that cpu time.

[kozumo]

Ig0r: Ipchains had already been compiled in the kernel and the biggest log I found in /var/log/ was 200 kb and it wasn't from portsentry. I checked if any portsentry files were unusually large, but none of them were.

I compiled the kernel again to see if I had done something wrong in the Network sections and I don't think I did.

Xsecrets: Definitely Portsentry and Syslogd. They're the first and second on the 'top' list respectively.

I killed portsentry for now, but does anybody know what's my problem?

I was also wondering if I were using 2.4.5 shouldn't I be using IPtables instead of IPchains? I have both of them compiled in the kernel though, but which one should I be running?