Networkmanager service is shutdown too early

[Alan Cox]

On Fri, May 30, 2008 at 04:06:04PM -0400, Colin Walters wrote:
> implemented in userspace, but that doesn't mean it's just another daemon,
> any more than TCP stacks that live in userspace in some old operating system
> designs could just be restarted at any time without any obvious side effects
> on the system.

I would suggest you take a look at the failover work people have done in
user space and also the higher level communication protocols. There is
a reason it is done higher up the stack - you want to deal at transaction
level.

I can *turn off* my NFS server, go for lunch, come back, turn it on and
my client -keeps working-.

I can restart the old X font server live

I can restart all my web services while using them and I maybe get a 
reset or odd error then it recovers

I can unplug and replug hard disks and it'll recover nowdays in almost all cases

Software should try and limp on and/or recover itself nicely. That way you'll
get a usable product that degrades acceptably. You wouldn't build a house
that fell down if any single brick cracked.

> Speaking of userspace kernel integration, I'd be pretty surprised if you
> could reliably stop the fuse service too; given all the discussion about
> nearly unfixable race condititons in rmmod that have (
> http://lwn.net/Articles/31474/).

Fuse is pretty close and its considered a bug that it isn't 100% perfect
on this yet.

> Auditing every program and making them able to cleanly support restarts of
> system services like HAL, NetworkManager, and PackageKit would be a huge
> task, for similar reasons because these services build up state. It might

It would be a huge step forward if they firstly just restarted as a default
response to dbus went away. Almost every one of them builds up a new state
when run from startup and it'll be far better than "oh reboot again" Windows
mentality crap. Some like NetworkManager need to be a bit smarter - but not
a lot.

> sort of work most of the time, but what you *really* do not want is for the
> system to sometimes break when the user plugs in a USB key when the
> background package upgrade happened to restart the bus at just the wrong
> time.

Most of the time is better than not restarting which means it *never* works.

Which do you prefer

	Insert USB key - nothing happens
	Repeat 50 times, reboot

or

	Insert USB key - nothing happens
	Repeat - ah success

Aaln