nullfs stabilization I
dillon at apollo.backplane.com
Thu Jan 12 10:55:56 PST 2006
: Although I didn't want do it, from a coding perspective we can make
: this glue sticky... that is, we can disallow the destruction of
: a lower layer namecache record while a higher layer namecache record
: in the same chain exists. This would remove the requirement that
: lower layers do callbacks to upper layers.
Er. This is slightly confusing. What I mean here is that the hard
part of coding a callback would be removed. Lower layers would still
have to be able to invalidate higher layers (i.e. rename performed
on the base filesystem), but now they would only have to follow the
preexisting chain upwards to do so and only deal with elements that
already exist, so no callback would need to be done. The namecache
code would be able to do it internally.
Also note that the idea of a 'shared locking structure' for the
shadow chain is very, very close to the fact that all the namecache
records will share the same vnode. The difference of course is
that the vnode might not always be resolved, hence we cannot embed
the lock in the vnode and instead need an auxillary structure that
the namecache records in the chain reference which contains the lock.
Another note: 'invalidating' a namecache record in DragonFly does not
involve destroying the structure. It is simply a matter of setting
the state in the structure, so we don't get into deadlock trouble
if other entities have active references to structures at various
points in the shadow chain.
More information about the Kernel