Re: extended attributes

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Fri, Jun 10, 2011 at 03:52:17PM +0000, Emmanuel Dreyfus wrote:
> On Fri, Jun 10, 2011 at 09:30:40AM -0500, Eric Haszlakiewicz wrote:
> > eh?  Don't corrupted filesystems cause a panic anyway?  What is it about
> > extended attributes that makes this more likely?
> 
> You rarely write to the executable files that you need in order to boot
> to single user, but modifying attribute backing stores is a common
> operation. This makes the later more susceptible to being corrupted at
> the time the filesystem is mounted.

I don't really understand.  Modifying the attribute backing stores is
subject to the same constraints as any other metadata write -- isn't it?

Is there somehow less protection against corruption here than there
would be when, for example, modifying the _inode_ for a given executable?

Shouldn't that operation be considerably _more_ common than modifying
the attributes for a given executable?

Thor