On 23/11/22 17:51, Karel Zak wrote:
Hi Ian,
On Wed, Nov 16, 2022 at 10:00:39AM +0800, Ian Kent wrote:
There's further discussion of this in the first patch.
I started trying to do this quite a while ago, glibc, automount(8) and
the kernel have been updated to use such a mount option but, for some
reason, when I canme to update util-linux I thought it would cause
problems for systemd. But when I looked at what I had done just recently
I saw that isn't the case so I'm now continuing with this change.
It's definitely no problem if the new behavior is optional and it's
possible to enable/disable it. So, the systemd can keep it disabled
;-)
Indeed, agreed, ;)
What I've done might not be the prefered way to do this so any comments
are welcome so it can be done the way it fits best with util-linux.
What about findmnt(8), maybe we can add --autofs=yes/not.
I think those that want this most will expect it to just work.
Typically they will be system admins that want to get rid of mount table
clutter resulting from autofs. So just changing an autofs configuration
option and having this work is the goal.
Historically autofs file system mounts didn't show up in mount table
listings and that's essentially the behavior I'm trying to bring back.
The problem is that we probably do not want to change the default
output as it's probably already used in many scripts, or is it
acceptable to hide autofs by default? It's already done for glibc, so
it may not be a big issue. Not sure.
Of course yes, and that's also the main reason we can't filter the
proc file system directly from the kernel, it must remain the
authoritative source with all mounts present so they can be seen if
needed.
In order for this to work the autofs user (typically a system admin)
needs to change an autofs configuration setting to make automount(8)
include the "ignore" option with autofs mounts so it's a conscious
choice. Otherwise there's no change.
Ian
