Re: 2.6.6-rc2-bk3 (and earlier?) mount problem (?

[Grzegorz Kulewski <kangur@polcom.net>]

On Tue, 27 Apr 2004 viro@parcelfarce.linux.theplanet.co.uk wrote:

> On Wed, Apr 28, 2004 at 12:04:59AM +0400, Paul P Komkoff Jr wrote:
> > > Excuse me?  The damn thing had found nothing.  However, it didn't care
> > > to release the devices it had claimed - hadn't even closed them, as the
> > > matter of fact.  That's a clear and obvious bug, regardless of any
> > > disagreements.
> > 
> > As far as I can see from here, evms parsed partition table, called
> > dmsetup several times and created corresponding nodes in /dev/evms.
> 
> ... without saying anything?
> 
> > Logic is easy - evms trying to concentrate block device management
> > into its own hands, but we have in-kernel partitioning code to
> > consider ...
> 
> How nice of them.
> 
> Well, AFAICS that means
> 	a) either kernel side of the things or the userland tools should
> printk/syslog - at least that evms device had been set up
> 	b) any distribution that runs this from initrd/init scripts would
> better take care of having sane fstab.
> 	c) nobody sane should put that as default.  Oh, wait, it's gentoo
> we are talking about?  Nevermind, then.

But what default? Gentoo just calls evms_activate before mounting 
filesystems to check if there are evms volumes (because filesystems can 
reside on it). And, according to man page, this is the right usage of 
evms_activate. And Gentoo could not know easily that I have no evms 
volumes in the system. And I possibly could insert some and boot Gentoo 
(and for example have label not device in fstab to specify fs to mount). I 
really think that evms_activate should not take the device that is not for 
it. And I think Gentoo is right. (And it is the best distro I ever tried.)


Grzegorz Kulewski