Re: [Fedora-livecd-list] "Input/output error" and/or "Bus error" for most commands

>  From: James Heather <j.heather(a)surrey.ac.uk&gt;
>  To: <livecd(a)lists.fedoraproject.org&gt;
>  Date: 03/11/2013 09:37
>  Subject: Re: [Fedora-livecd-list] "Input/output error" and/or "Bus 
>  error" for most commands
>  Sent by: livecd-bounces(a)lists.fedoraproject.org
>  
>  > Critical limitation
>  > LiveOS overlays are employed as write-once files, which always consume
>  > space—never releasing space for reuse. Should the overlay storage
>  > space, whether temporary or persistent, be totally consumed, the
>  > system will likely crash with Input/output or Bus errors.  If such a
>  > crash does occur while using temporary storage space for the overlay,
>  > a simple reboot will rectify the situation.  With persistent storage
>  > the situation is more dire and will require ... (I don't know this
>  > case at all).  In either case, achieving the reboot likely will
>  > require a hard reset since attempting a software initiated reboot will
>  > probably fail with more Input/output or Bus errors. 
>  
>  ... will require appending "reset_overlay" (think that's right) to the
>  kernel command line on boot-up, which will reset the state to its
>  initial boot (though it won't reset a persistent home, of course).
>  

I've updated the page with this much and confirmed/noted that 
"reset_overlay" is the former (but now deprecated) form of 
"rd.live.overlay.reset".  Please review/adjust as necessary.

 
>  It's not quite true that a live overlay will never release space.
> I
>  think if you created a file that got stored in the overlay, and then you
>  deleted it, you'd reclaim the space. But what it can't do is reclaim
>  space that's used by the squashfs image: deleting things from there will
>  do something to mark them as deleted in the filesystem, but you won't be
>  able to use the space, since it wasn't ever taking up any overlay space
>  in the first place.
>  
>  I think the overlay operates at the sector level, but I'm not certain of
>  that.
>  
>  James
>  
I left this bit out.  It's good info, but sounds like a bit of certainty 
could be injected.  ;-)

Also along these lines, a coworker was playing with the 'discard' mount 
option to see if the kernel's TRIM support would help here.  Testing 
showed that it does not and further research discovered kernel mailing 
list comments stating clearly that this feature is not yet available for 
the COW images.

--
John Florian