Port update hosed entire system
ateve at sohara.org
Tue Oct 2 08:31:28 UTC 2012
On Tue, 2 Oct 2012 09:16:43 +0200
Polytropon <freebsd at edvax.de> wrote:
> On Tue, 2 Oct 2012 06:20:45 -0400, Rod Person wrote:
> > It would never have occured to me that updating a port that
> > has to do with audio and video containers would totally leave me unable
> > to login into my system or issue and shell commands without getting
> > a segmentation fault.
> I find it very hard to see a correlation here. Coincidence? Yes,
> but I cannot imagine a way a port can dmage the system in that
> way so not even shell commands keep working...
> > I did discover that my / file system had run out of space -131MB.
> That could show that some part of important content on / has
> not been written yet - it's still held in "write buffers"
No, the negative free space simply means that you have encroached
on to the reserved space (only root can do this) which is usually used to
optimise the layout when writing new data.
> pending. So you could first check what takes up space in /
> that is not required to be there, and remove it, then the
> "write buffers" will be written properly. A "sync" command
> could do this on request.
Having negative free space will prevent non root users from writing
data, but that will be returned to the applications as error returns to
write calls not held in write buffers.
> Check with "df -h" for _no_ negative values before rebooting
> the system into SUM. I'm not sure if the "write buffers" can
> survive a shutdown.
They can't but they're not connected with negative free space
reports. A normal shutdown will flush all the buffers.
> > I'm still able to issue sudo, so using sudo rm -r I was able to free up
> > 25GB...but still, /bin/sh, ls, clear all seg fault and su doesn't work
> > and switching consoles doesn't let me log in.
> That sounds that somehow calling programs (executing / forking)
> is not working properly anymore. As this is one of the most
> fundamental mechanisms of the systems, it's hard to believe
> that this can be triggered through a port update...
More likely one of the shared libraries they all use has been
overwritten. Updating ports certainly shouldn't be able to do this though.
The stuff in /rescue should work fine for getting a usable
environment to go bug hunting in, but without a deep and intimate knowledge
of how things are supposed to be it's going to be hard short of
Steve O'Hara-Smith <ateve at sohara.org>
More information about the freebsd-questions