[GRLUG] Interaction: NFS and mount loop

[George (Skip) VerDuin]

I ran into something interesting this weekend.
What I think I *may* have learned is this:
   When an .iso image is mounted, 
     the internal contents of the .iso may *NOT* be accessed via NFS.
That was a surprise...

Here is what I am up to.  I have [two now, more later] old PCs to load
with a custom application for ham radio use as "remote" packet radio
station controllers.  I am searching for a good platform; FC-8, ubuntu
7.10, DSL, debian, have been considered.  DSL so far is the winner, but
I want to give debian a few more tries.

The raw machines don't have DVDs but they do have NICs so to install
FC-8 I set up FC to load via NFS.  I already have the FC-8...iso I used
for my server [update from hard drive option], so my first choice was to
mount the iso and point the raw machine into the mounted directory via
NFS for access to the material.  NOPE.  I had access to all directories
in the path except the last directory that I needed.  The NFS error
message refused to "mount the specified directory".

SO - I extracted [copied] the material from the .iso into it's own
directory structure.  This action did allow access to the distribution
material and the FC-8 install proceeded via NFS nicely.  Later the
install failed because of lack of space on the raw machine -- but at
least the install process worked.

As far as the installer is concerned, there is a difference between:
 a) a directory containing distribution material
 b) a directory mounted from a .iso containing distribution material
Now as I re-think the experience, I did not attempt a third option --
 c) a directory containing the .iso file containing dist. material
Excluding the c) option, I can swear to trying lots of parameter
varieties relative to the b) option all without joy.

To extract the .iso material, I did:
...# mount -o loop,ro -t iso9660 dl/file.iso spool
...# cp -a spool spool2
What I find to be interesting is that all commands from the command line
have equal access to both spool and spool2, and that NFS could serve
from spool2 but not from spool.  What it means to me is that:
  A) "NFS" and "mount -o loop" have a bad interaction
  B) I messed up the parameter set for my work

Does anyone have further insight into my experience?  Did I use wrong
parms on the mount or on NFS?  Is installer really excluded from access
to the contents of a .iso image by way NFS after mounting it?

After the FC experience, I did install debian via NFS from a mirror
somewhere on Internet starting from a floppy disk boot.  Presently I am
loading a mirror on my LAN to avoid bothering the Internet mirror again
as I try various configurations.

In any case -- If installer "should" be able to access a directory
mounted from a .iso image, I think I will report this to FC as a bug...

73
de [George (Skip) VerDuin] K8RRA k