[RFC,PATCH 0/4] Dynamic Pseudo Root

[J. Bruce Fields]

On Mon, Dec 10, 2007 at 10:01:11AM -0600, Tom Haynes wrote:
> J. Bruce Fields wrote:
>>
>>   
>>> Finally, how do you store the state?
>>>     
>>
>> Which state?
>>
>> --b.
>>
>>   
>
>
> Whether or not the export has been loaded or not.

The kernel literally has no record of the export until it's used.  It
does know about mountpoints, of course, but doesn't know which mounted
filesystems are exported, who they're exported to, or with what options.

> Although it sounds like from your other answers that you could check to
> see if it is in memory from the info from the FH.

Yep, that's right.

> Which goes back to how do you detect if the filesystem in question has
> not been exported yet or is offline for administrative purposes?
> I guess you try to export it and detect that it is offline.
> And then do you want to cache that information or just keep on
> checking each time?

Apologies, I can't remember off the top of my head how we handle that
case (or if we do!).  We should allow a filesystem that knows some media
is absent (for example) to return an error that gets translated to
DELAY/JUKEBOX, but we don't have any explicit support that I know of for
marking an export as temporarily unavailable.  If somebody requests it,
then I think it would make sense to add an extra export flag and either
pass down the export with a short expiry (there is an expiry time
associated with each), or leave it to mountd to notice the change.

--b.