mystery /proc and spamhippo

[Sheer Panic <sheer eni net>]

I have two  unusual problems. The first is more or less the usual unusual,
the second is well, just, odd.

The first problem is I have a axp-linux box (533Mhz LX164, 12 4g disks,
3 duel channel intraservers) running innd. All is well with it except that
since I installed the spamhippo patch to innd, I get unaligned accesses
left and right. (about one for every incoming message) has anyone worked
with spamhippo (www.spamhippo.com) and innd on axp-linux? If not, does
anyone have any suggestions for what in the code I might want to look for
as causing unaligned accesses?

The second problem is I have an axp-linux box (533Mhz, LX164, 8 4g disks,
2 duel channel intraservers) that every once in a while, just for the heck
of it, decides that syslogd and klogd need to go into overdrive.. and I
mean OVERDRIVE. They consume approximately half of the CPU each. I notice
that in my messages file before this occurs:

Mar 27 04:31:44 mail4 sshd[3939]: log: ROOT LOGIN as 'root' from
chastity.eni.net
Mar 27 04:31:44 mail4 kernel: Cannot read proc file system: 9 - Bad file
descriptor.
Mar 27 04:32:15 mail4 last message repeated 300159 times
Mar 27 04:33:16 mail4 last message repeated 636616 times
Mar 27 04:33:21 mail4 last message repeated 43818 times
Mar 27 04:33:21 mail4 kernel: Kernel logging (proc) stopped.
Mar 27 04:33:21 mail4 kernel: Kernel log daemon terminating.
Mar 27 04:33:26 mail4 syslogd: exiting on signal 15

[root@mail4 /root]# ls /proc
1            245          3941         7            ksyms        scsi
1100         246          3978         cmdline      loadavg      self
171          247          4            cpuinfo      locks        stat
182          248          421          devices      mdstat       sys
193          249          433          dma          meminfo      uptime
2            250          442          filesystems  misc         version
204          252          471          interrupts   modules
217          253          478          ioports      mounts
229          3            5            kcore        net
242          3939         6            kmsg         pci
[root@mail4 /root]# 

As you can see, the kernel was entering this message approximately 600,000
times a second. Okay, I can see where this could cause a little system
slowdown.. but WHY?

I look into the kernel source and..oh my god there are GOTO commands in
the linux source. But something tells me this isn't what's going on..

Thinking about it, I suspect klogd somehow got wedged into generating
messages as fast as it could. Does anyone have the klogd source handy?

S.