Public SourceMage Discussion List

[Seth Alan Woolley <seth AT positivism.org>]

Actually, it's not for ext2 (it's safe there), it's for other 
filesystems.  There's no "vote" issue, as SOFTFIX=yes should be default 
for now until we can reimplement the crappy fsck command.

It's apparent from other replies (I'll just comment here since I don't 
feel like starting a new email) there's some confusion on what fsck 
does... it just passes checks and unrecognized arguments onto 
filesystem-specific checkers "in a particular order" depending on what's 
in fstab.  The problem is that filesystem-specific checkers are doing a 
really bad job of using consistent switches.  So while the fsck manual 
says -a is dangerous for some, everybody uses it anyways since ext2 maps 
it to -n.

I just think we need a lot more intelligent fs checking on startup 
because if an error happens bad stuff happens.  I'd almost suggest we 
spawn fsck checking off onto another unused terminal with an open tty 
and use IPC (perhaps a file semaphore) to let init know it's done (and 
deallocate the vc when it's done).

That way the fscks can never claim there's no terminal open (which is 
causing most of the headache), plus you can do a bit more verbosity with 
it since it's outputting to another terminal (perhaps 11 like how 12 is 
often the logger).  I'm also thinking perhaps we can add another field 
to fstab to let the system know which filesystems are required for 
continued booting.  Everybody hates it when a fs fails and it's not 
really needed to boot but it croaks the whole boot process.

Anything to make boot more graceful/intelligent is better, I think, as 
there's been a historical lack of creativity wth the fsck process in 
unix/linux as far as I'm aware, and it really needs some help, at least 
the linux fsck.

Seth

On Sun, Jul 24, 2005 at 02:21:53PM -0400, Sergey A. Lipnevich wrote:
> So what you're saying is that with `-a' fsck is not as good as with
> `-n', at least for ext file system? I don't use ext{2,3} and I don't
> remember being in a situation where I actually needed fsck to run with
> any other settings except default, even after power outage or forced
> power-off to restart a frozen machine. I'd vote for SOFTFIX=yes being a
> default.
> Thanks!
> 
> Sergey.
> 
> On Sat, 2005-07-23 at 10:00 -0700, Seth Alan Woolley wrote:
> > It probably should be the default until all the problems with 
> > filesystems we support are resolved with it.  I think the only way will 
> > be to run fsck manually on filesystems that load instead of using fsck 
> > directly to indirectly call fs-specific fscks.
> > 
> > I'll have to investigate each filesystem empirically as the 
> > documentation for the various fscks on what they _actually_ do is 
> > severely lacking and the man pages seem to be wrong on many accounts on 
> > what works and doesn't. :( I rather thrash my filesystems and when they 
> > have problems (usually caused by a power outage during massive 
> > thrashing) the fsck at boot has been too liberal for what I desire, 
> > which is why I applied the change in devel in the first place.
> 

-- 
Seth Alan Woolley [seth at positivism.org], SPAM/UCE is unauthorized
Quality Assurance Team Leader & Security Team: Source Mage GNU/linux
Linux so advanced, it may as well be magic http://www.sourcemage.org
Secretary Pacific Green Party of Oregon http://www.pacificgreens.org
Key id FDCEE733 = 5302 B414 64C4 6112 3454  E082 99F0 69DC FDCE E733

Attachment: pgpE8bkG8k04Y.pgp
Description: PGP signature