> > I assume this happens because sysklogd reads the messages from
> > the scrollback buffer after it is started, in order to log
> > messages which have gone by already. It just isn't reading back
> > far enough, as the info is there in the scrollback buffer
> > (I can see it).
>
> Actually, it's an internal buffer. See kernel/printk.c. It's
> 8k.
Sort of a semantic point.
> If you're generating more than 8k of data, perhaps you can
> rearrange things to start up syslog sooner.
Or the buffer could be extended. 8k isn't much, this might make a nice
config option if it isn't already.
lilo