RFC: deprecate mkfs.xfs protofile functionality

RFC: deprecate mkfs.xfs protofile functionality

[Eric Sandeen]

I'd like to deprecate and eventually remove the protofile functionality
in mkfs.xfs, because it's not well tested code which is really a special
snowflake in the codebase that adds complexity.  And as far as I know,
nobody uses it.

If I'm wrong though, speak up now.  Is anybody using protofiles?

(If you don't know what a protofile is, then you're not using it.
And let's not turn this into a "well I could think of a way to use it"
discussion, because it's been there for a couple decades and if you're
not using it /yet/ ... unless there really is some new usecase on the
horizon which would clearly depend on it.)

Thanks,
-Eric

Re: RFC: deprecate mkfs.xfs protofile functionality

[Darrick J. Wong]

On Thu, Dec 20, 2018 at 08:34:25AM -0600, Eric Sandeen wrote:
> I'd like to deprecate and eventually remove the protofile functionality
> in mkfs.xfs, because it's not well tested code which is really a special
> snowflake in the codebase that adds complexity.  And as far as I know,
> nobody uses it.

It also would seem to barf out corrupt long format symlinks on v5
filesystems, as I discovered while cleaning up mkfs and repair
yesterday.

> If I'm wrong though, speak up now.  Is anybody using protofiles?
> 
> (If you don't know what a protofile is, then you're not using it.
> And let's not turn this into a "well I could think of a way to use it"
> discussion, because it's been there for a couple decades and if you're
> not using it /yet/ ... unless there really is some new usecase on the
> horizon which would clearly depend on it.)

I'd rather have a mkfs feature that copies everything under a given path
into a newly formatted filesystem, and even then only if there's an
especially good reason why mount/cp/umount cannot be done.

So I say nuke the protofiles in xfsprogs 5.0. :P

--D

> Thanks,
> -Eric