Re: fallocate for FFS anyone working on this ?

[David Holland <dholland-tech%netbsd.org@localhost>]

On Sat, Mar 17, 2018 at 01:16:38PM +0530, Manas Mangaonkar wrote:
 > I am a computer Engineering Sophomore year student and want to work on
 > this(not as a part of GSOC or anything else) Is anyone else working on
 > this,If no can you point to some resources that will help me get started.I
 > want to begin coding asap. Familiar with Unix

Nobody is working on it as far as I know.

A good place to start (if you haven't already) are the FFS papers,
which are old but considered classics. There are copies in
/usr/share/doc (mckusick-84 and mckusick-85) -- also compare what you
find described there to the current on-disk data structures that you
can find in the headers in src/ufs/ufs and src/ufs/ffs.

Unfortunately there's no paper on the journaling mode we have
("wapbl"); there's articles on journaling for ffs, but they're about a
different implementation.

Basically what fallocate is supposed to do is fill in holes in a file
with freshly allocated blocks, without violating the recovery
invariants assumed by fsck (and wapbl's journal replay). This is
mostly a matter of allocating blocks and zeroing them out, and fitting
that into everything else that's going on.

Arranging to not have to zero the new blocks explicitly would be nice
but is quite a bit harder.

-- 
David A. Holland
dholland%netbsd.org@localhost