I have a UDB with a 2.1.78 kernel. Occasionnaly, syslogd becomes wild
and dumps countless messages "Exception at ..." that bring the box to
its knees.
Today I had this while trying a newly compiled portmap, and this was by
rlogin. The system seemed to crash, so I unlogued from my Sun station
and came to see my UDB... and it was ok. It seems that my unlogin
killed the portmap that was causing the messages (it was in debug mode
and so lost its terminal when I unlogued) and the messages stopped.
So... this time I _have_ the messages. There are two of them, alterning :
Jan 12 19:40:22 orion kernel: portmap: Exception at [<fffffc00003b15bc>] (fffffc00003b15c4)
Jan 12 19:40:22 orion kernel: portmap: Exception at [<fffffc00003f4e80>] (fffffc00003f4efc)
The first one is in write_chan() in drivers/char/n_tty.c
The second one is in __copy_user, in arch/alpha/lib/copy_user.S
I think about some sort of exception loop, where the treatment of an exception
(printing it on the screen) triggers another identical exception.
I saw such messages also under intensive harddisk work or network activity.
Switching off syslogd somehow cures the problem (the box does not crash and
remains usable) but this is not satisfying (I _want_ my syslogd).
By the way, if someone could indicate me where I can find a running portmap
daemon...
--Thomas Pornin