Re: DiskDevice API v2.1

Tyler Dauwalder <tyler-cyCl/[email protected]> wrote:
> On 2003-04-08 at 12:01:21 [-0700], Ingo Weinhold wrote:
> > Tyler Dauwalder <tyler-cyCl/[email protected]> wrote:
[...]
> > > > To store the data on the boot partition is an option, I think. 
> > > > I
> > > > wouldn't use a file, but leave some room for the meta data at 
> > > > the
> > > > beginning. ReiserFS for instance skips the first 64 KB. That 
> > > > would
> > > > be
> > > > space enough for quite a bunch of RAID settings and a small job
> > > > journal. But even 1 MB or so wouldn't hurt. Perfect would be to
> > > > make
> > > > this an option when initializing the volume. Be's BFS add-on
> > > > wouldn't
> > > > be able to read such a volume, but I think, that doesn't matter
> > > > much.
> > > 
> > > That seems like a good idea. Though why not in a file: the access
> > > overhead, or contiguity, or...? If it were a setting at 
> > > initialize
> > > time, you could certainly guarantee a contiguous chunk.
> > 
> > Firstly, I wouldn't see the benefit of a file, if you have to 
> > create 
> > it
> > at initialize time and can't change its size later on. 
> 
> Doesn't hurt compatibility w/ BFS.

Ah, right.

> > Secondly, there
> > is a good chance, that you have to move that file, when you're 
> > resizing
> > the partition. If you place it at the beginning of the data space, 
> > you
> > might have to move it, when the partition is enlarged (the bitmap 
> > will
> > grow). And otherwise you might have to move it, when the partition 
> > is
> > shrunk.
> 
> That could be handled gracefully (and contiguously) as well, though.

Sounds more complicated though. In particular because the interaction 
between the system (kernel, disk device manager -- whoever is 
responsible for partitioning) and the FS will be more complex, since 
the FS would move the log file, which itself is a transaction that 
needs to be logged, and while the system is responsible for logging. I 
guess, we need to work on the details before being able to do further 
evaluation. It may be worthwhile to pursue, though.

[...]

CU, Ingo

PS: It looks like I won't finish writing up the kernel ideas today 
either. I had less time, than I was hoping. *sigh*