submount?

Alan Cox alan at redhat.com
Mon May 31 19:30:12 UTC 2004


On Mon, May 31, 2004 at 09:18:40PM +0200, David Zeuthen wrote:
> > If it works sure. Ultimately I'd like to find a clean way to do a volumagic
> > like thing with the I/O layer in kernel space. Let the daemon do handle
> > translations and let the kernel do the I/O paths
> 
> Wouldn't it suffice with just an option to mount(1), e.g. have as few
> changes to how this impact user space as possible. I mean, at the end of
> the day, isn't this just about the kernel locking/unlocking the drive
> (same for e.g. USB storage) only when there is I/O?

To do it properly you need a bit more because users may change CD-ROMs and you
probably want to do "Please insert 'Install Disk 2'" type dialogs. 

Internally the kernel really doesn't deal with disks going away when they are
wanted. Volumagic can cope with

	cd /mnt/cdrom
	eject
	insert new cd
	ls

and also with requests when the disk is removed - not all file systems handle
media vanishing as a routine thing.

Stephen Tweedie had some 2.2 patches "supermount" which tried to address this
whole thing properly and while there have been 2.4 and 2.6 attempts I've not
seen one that addresses it all cleanly. 





More information about the fedora-devel-list mailing list