Re: [RFC][PATCH] rbind across namespaces

[Mike Waychison <mikew@google.com>]

Miklos Szeredi wrote:
>>>I still see a problem: what if old_nd.mnt is already detached, and
>>>bind is non-recursive.  Now it fails with EINVAL, though it used to
>>>work (and I think is very useful).
>>
>>Hey, you just made another argument for not detaching mounts when the
>>last task with that current->namespace exits, but instead detaching
>>mounts when the last reference to any vfsmnt in the namespace is dropped.
>>
>>Hint :)
> 
> 
> I have a better idea:
> 
>  - create a "dead_mounts" namespace.
>  - chain each detached mount's ->mnt_list on dead_mounts->list
>  - set mnt_namespace to dead_mounts
>  - export the list via proc through the usual mount list interface
> 
> The last would be a nice bonus: I've always wanted to see the list of
> detached, but not-yet destroyed mounts.
> 
> Does anybody see a problem with that?
> 

FWIW, all this stuff has already been done and posted here.

Detachable chunks of vfsmounts:
http://marc.theaimsgroup.com/?l=linux-fsdevel&m=109872862003192&w=2

'Soft' reference counts for manipulating vfsmounts without pinning them 
down:
http://marc.theaimsgroup.com/?l=linux-kernel&m=109872797030644&w=2

Referencing vfsmounts in userspace using a file descriptor:
http://marc.theaimsgroup.com/?l=linux-kernel&m=109871948812782&w=2

walking mountpoints in userspace: 
http://marc.theaimsgroup.com/?l=linux-kernel&m=109875012510262&w=2

attaching mountpoints in userspace:
http://marc.theaimsgroup.com/?l=linux-fsdevel&m=109875063100111&w=2

detaching mountpoints in userspace:
http://marc.theaimsgroup.com/?l=linux-kernel&m=109880051800963&w=2

getting info from a vfsmount:
http://marc.theaimsgroup.com/?l=linux-kernel&m=109875135030473&w=2

Mike Waychison