bug-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Ext2fs 2GB limit status?


From: Marco Gerards
Subject: Re: Ext2fs 2GB limit status?
Date: 21 Sep 2003 20:15:07 +0200
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2

Ognyan Kulev <ogi@fmi.uni-sofia.bg> writes:

> Marco Gerards wrote:
> > Ok. Perhaps it is interesting to discuss journaling.  AFAIK Linux has
> > some journaling code shared by all filesystems.
> 
> It doesn't.  There is an "universal" JBD layer that is designed to be
> used by any filesystem, but currently only ext3 uses it.

Oh, that's what I meant.  Ofcourse it doesn't for me if it is used on
Linux, what matters more is the question "Why isn't it used?".
Perhaps there are problems in its design, stuff we should avoid.
 
> > Perhaps it is useful to extend libdiskfs with such feature.
> 
> I'm still on the block layer, but changing libdiskfs to support
> journalling seems almost inevitable for ext3fs support at first glance.

Ok. :)

> > Perhaps you can share some
> > code for removing the 2GB limit too?  I think it is a waste of time to
> > fix the limit for every individual filesystem.
> 
> Actually, fixing the 2G limit was just a step towards ext3fs for me.
> It helped me a lot to go deeper in ext2 and the Hurd.  In ext3fs,
> there will be much more changes in the block layer, which (if core
> developers like it) will be integrated in ext2fs.  I even think about
> integrating some kind of soft deps, like in BSDs.  But for now, it is
> much more important to have _some_ ext2fs without this stupid limit
> that makes others laugh at us.  Unfortunately, ext3fs must be finished
> ASAP for personal reasons.

I agree. If you want to hear my opinion or want me to look at some
code, just kick my ass. :)

> > I'm even thinking about how easy it is, if this is really general, to
> > add journaling and ordered writes to filesystems that don't even
> > officially support this by just having a nice journaling
> > layer. (This is just a random thought that is possibly quite useless).
> 
> That's the idea of JBD, but noone uses it in this way yet.  It's not
> an easy job to journal random filesystem with JBD, though.

The filesystem must provide a way to determine how the meta-data is
related and in which order it should be written.  Ofcourse it is not
possible to use a journal in the fs, but you can provide an external
journal.

Thinking of fatfs, I think such thing is not that hard to implement,
same for stuff like minixfs, etc.

Thanks,
Marco






reply via email to

[Prev in Thread] Current Thread [Next in Thread]