>last weekend, the backup of our biggest HSM filesystem (AIX 4.3.2,
>ADSM 3.1.0.8) expired all files !!!

Reinhard - My sympathies on that ugly problem.  It superficially sounds
           as though the file system was not mounted, such that the
backup found just the empty mount point directory and worked that.
I would hope it not the case that HSM asked for file system info for
the backup and, because of a file system problem at the time, got none,
and kept going.  The backup log from that time may provide clues.
The EINTR errno on statfs() is not a documented errno for that system
call, so its meaning in that context is undefined.  I have not experienced
it in my experience with programs utilizing statfs().

Looks like you're stuck having to perform a complete backup again.
Something peculiar transpired in your system during that time.
Perhaps the AIX Error Log will have info on it, or the dsmreconcile
or dsmautomig logs will reflect some conditions.
If no AIX reboot occurred since that time, I wonder if vestiges of
the condition linger, and may be affecting your backup processing now.

>- The new backup process is now recalling all files, although I expected
>  it to simply make a copy within the TSM server. Why? ("Migration
>  Requires Backup?" is set to "No".)

With the HSM and backup storage pools within the same server, backup
is *supposed* to copy the data within the server, without recalling
it to the client file system.  (If the data has not yet been migrated,
or is premigrated, backup will occur from the filesystem copy of the
file.)  Files marked for read-without-recall will be backed up through
the client in that manner; but files have to be individually marked
that way, and I doubt that it applies in your case.

I'm wondering if perhaps the client options got changed or a different
server stanza is being referenced to confuse the process?
Wish I could offer more specifics,
    Richard Sims, BU