server hanges at 100% cpu with connection attempts

[Mikael Ostensson]

The reason for glibc is that we are going to run apps built with
freepasacal on the system (porting windows apps written in delphi) and
it seem to not be any easy way to link to uclibc even thought that
might have been prefered. I hope my setup of glibs is enough though.

After reading abit over the weekend the problem might be that I need
to mount the pts file system under /dev. I'll try you suggestions an
let you now how it works out

Mikael

On 11/19/06, Matt Johnston <matt at ucc.asn.au> wrote:
> On Fri, Nov 17, 2006 at 07:05:47PM -0500, Mikael Ostensson wrote:
> > I have started from the begining: Dowloaded the latest kernel source, build
> > a kernel and initfs image with busybox.
> > busybox is complied dynamically against glibc (the version that comes with
> > fedora core 6, my build box belive it's 2.5 from the numbers listed in the
> > libs).
>
> Is there a reason you're using glibc rather than uClibc? In
> my experience uClibc is a lot less painful for setting up an
> embedded-style system from scratch. (Though if it works, you
> may as well stick with glibc).
>
> > I can run dbclient to my host box fine, but when connecting to my build box
> > the spawned process goes 100% cpu and wont return untill killed.
> >
> > Reading the list archive I atempted connecting with : ssh -t user at IP "echo
> > 123" to see if the was a problem creating tty/pty  but the result is the
> > same.
>
> Giving -t will explicitly allocate a tty/pty. Try with -T
> instead and see if that works.
>
> Can you run gdb or strace on the system? If you run "strace -p <pid>"
> it'll give you a list of syscalls, or:
>
> gdb --pid=<pid>
>             ... gdb loads
> bt          (backtrace)
>
> will show a stack trace where it's spinning. If you continue
> ("c") then break (ctrl-c) and "bt" a few times you should be
> able to statistically figure where the problem is.
>
> Let me know how that goes.
>
> Matt
>