On 03/06/07, Gary Duzan <gary@duzan.org> wrote:
>    I recently switched my amd64/current box to a new disk and set
> it up with an LFS root file system. Since then I've had the system
> hang occasionally and become mostly non-responsive. If not in X,
> I can switch consoles and get into the debugger, but that's about
> it.
>
>    I can reliably recreate the problem by trying to rebuild firefox:
>
> ###########################################################################
>
> % cd /usr/pkgsrc/www/firefox ; make clean && make && make update
>
> ###########################################################################
>
> It finishes cleaning and building, but hangs in pkg_delete when
> updating, as seen in the debugger's process listing, included below.
> Unfortunately, I can't seem to get a stack trace for anything in
> the debugger; it only prints something like "trace: pid 249 lid 1
> at 0xffff80004b06a300".)
>

I used to have an IDE disk in lfs for quite a while without much
trouble; some times ago I moved some of the temporary directories used
during system builds - dest and obj - to it, as well as some of pkgsrc
infrastructure. This is an i386 system, tracking -current. I get very
similar hangs when there has been heavier than usual i/o to it.

I've switched back to ffs on this disk. I still have a couple of lfs
filesystems on an external 1394 disk, which I use mostly for
archiving, without any problems.

I couldn't get any sensible dump at the time. I haven't been brave
enough to setup a system with an lfs root.

>    Any ideas? This isn't a critical box, or I wouldn't be running
> current on it, but I'd like to help stabilize things if possible,
> and I'd be happy to help debug it.
>
>                                         Gary Duzan
>
>
>
>  PID           PPID     PGRP        UID S   FLAGS LWPS          COMMAND    WAIT
>  4379         12519      621          0 2  0x4000    1       pkg_delete seg_ioc
>  ...

Chavdar Ivanov