Re: Boycott Systemd

Frank Peters posted on Tue, 23 Sep 2014 10:55:58 -0400 as excerpted:

> Also, my example of the changes in USB device nodes is not the only
> recent occurrence of /dev tree modifications.  The kernel folks also
> removed the static /dev/rtc, or real-time clock device node.  In place
> there is now /dev/rtc1, /dev/rtc2, etc., and the intention is to
> dynamically allocate these nodes with udev.  This change broke my use
> space but it was easy to fix.

I guess you'd be the one to ask about this...

Have you tried the kernel's own devtmpfs?  How well does it work compared 
to a static dev, etc?

I run devtmpfs but with udev (and actually now full systemd) on top, so I 
don't know how it does by itself.  But best I know, the idea is that it 
dynamically handles the default kernel devices, popping them in place as 
the corresponding hardware is detected.  For general desktop systems udev 
is assumed to be run on top and do fancy stuff like the /dev/disk/by-* 
symlinks and specific non-default permissions (with default being root/
root 0660 IIRC), but for embedded and systems that have a pretty static 
device config and thus don't want/need the fancy udev stuff, devtmpfs is 
supposed to provide basic dynamic-device-node service.

So I'm wondering how well it works with your sort of config by itself, 
and whether it's a reasonable basic replacement for a static device 
tree.  It seems to me that might be the expected middle-road for those 
who don't want/need a full udev, and since it's a pure kernel and kconfig 
solution, including an option to have the kernel automount it without 
userspace help, that might be what they'd point you to as an answer to 
the otherwise userspace breakage.

But I haven't the foggiest whether devtmpfs would handle those dynamic USB 
device nodes without udev, or not.  My /guess/ would be that if it 
doesn't, making it do so might be the bug-fix they'd offer if someone 
/did/ complain about userspace breakage in that regard.

Like I said, you'd be the one to ask, so I am. =:^)

-- 
Duncan - List replies preferred.   No HTML msgs.
"Every nonfree program has a lord, a master --
and if you use the program, he is your master."  Richard Stallman