Re: fs weirdness during a reconfigure

We have never properly managed adding or removing cache directories. The
currect code may however be slighly worse than how it was when we only
had "ufs" (with/without async-io). If you can think of a good method you
are free to take a stab at it.

Adding:
  * When/how to call newfs?
  * When/how to call init?
  * What about cache rebuilds?

Removing:
  * Cleaning up the store index
  * What to do with locked/active objects?
  * And internal objects?

Changing their order is however no problem as the dir name is used as
key for identifying cache directories, and cache rebuilds ignores any
stored directory number when reading swap.state files (is the directory
number even stored nowdays?).

Regards
Henrik

Adrian Chadd wrote:
>
> Hi guys,
>
> There's something funky going on with the reconfigure code.
> Specifically, when new filesystems come into existance,
> their init function doesn't seem to be called.
>
> I don't know how many people actually add filesystems during a
> reconfigure, but even the ufs/aufs/diskd Init() routines aren't
> called.
>
> I know deleting a store or shifting their order is bound to cause
> massive confusion with the current code, but what about just adding?
> Shouldn't we at least print a warning or error out?
>
> Adrian
Received on Sat Oct 06 2001 - 02:40:57 MDT