mpd 4.1 on 5.5-STABLE

[Ian Smith]

On Sat, 3 Mar 2007, Alexander Motin wrote:
 > Ian Smith wrote:
 > > When started the first time, before there's any ng0 interface, mpd logs
 > > the following two lines then immediately exits without further ado:
 > > 
 > > paqi# /usr/local/etc/rc.d/mpd4.sh start
 > 
 > Try to run mpd from console just by typing "mpd4" to get all logs to 
 > console. To get more logs you can add to the begining of mpd.conf 
 > command "log +all".

Thanks for your prompt help Alexander; I've been at it most of today. 

That works fine.  What doesn't work here, very repeatably, is mpd4 -b

I can send some more details offlist if you like, but in a nutshell,
mpd4 -b syslogs the intro line, then sometimes another one, maybe two: 

Mar  4 20:03:16 paqi mpd: process 39879 started, version 4.1 (root at paqi.nimnet.asn.au 20:51  3-Mar-2007)
Mar  4 20:03:16 paqi mpd: CONSOLE: listening on 127.0.0.1 5005
Mar  4 20:03:16 paqi mpd: [b_PPPoE] exec: /sbin/ifconfig xe0 up

then just disappears; no (logged) message, no core.  Some other times it
keeps running but responds to neither telnet nor kill, short of kill -9.
Either way, it creates then leaves its netgraph stuff behind.

But when started in foreground without -b it's solid, and afaik so far,
working fine as indicated by a couple of test connections to $provider. 

Once started, ^Z,bg puts it in background where it's happy to run, also
if started with 'mpd4 -p /var/run/mpd4.pid >/dev/null &' so I've hacked
/usr/local/etc/rc.d/mpd4.sh to do that on 'start' .. also working well.

 > > 'ifconfig ng0 destroy' won't.  So then, now mpd will start, uselessly: 
 > 
 > ngX interfaces can be shutted down only from inside of netgraph by 
 > "ngctl shutdown ngX:" command.

The colon makes all the difference; ngctl(8) is kinda light on examples,
and I couldn't figure what <path> meant there till browsing netgraph(4)

 > >         new -i ng0 PPPoE PPPoE
 > 
 > -i parameter is not recommended for use. If you remove it, mpd will just 
 > use first free ngX interface.

I only read all the web docs twice, and trusted all the samples :)

Anyway, when not run with -b which reliably fails here, mpd4 runs and
cleans up after itself nicely, however stopped.  I'm now using
	new b_PPPoE l_PPPoE
which makes learning about the beast from its logs a bit easier for me.

 > > One thing makes me a bit nervous: those ipv6 and inet6 hooks.  kernel is
 > > compiled without INET6, new kernel/world built with make.conf having: 
 > > NOINET6=true	# do not build IPv6 related programs and libraries
 > > WITHOUT_IPV6=true
 > 
 > I don't think that this should be a problem.

Neither do I .. but there seems to be something different here, assuming
you haven't heard of -b mode failing reliably on other 5.5-S systems?

Happy to try anything else over the next few days, before it's up 24/7

Cheers, Ian

PS a bit of 'lastcomm -eE -f /var/account/acct.0 | grep mpd4' to show
how fast those -b attempts were exiting (except those that hung, FX): 

mpd4     -       root             ttyp0     1245.00 es Sun Mar 4 23:54
mpd4     -FX     root             __         285.00 es Sun Mar 4 23:28
mpd4     -       root             ttyp0        0.12 es Sun Mar 4 23:24
mpd4     -       root             ttyp0     1843.00 es Sun Mar 4 21:57
mpd4     -       root             ttyp0     1506.00 es Sun Mar 4 21:18
mpd4     -       root             ttyp0       60.36 es Sun Mar 4 20:16
mpd4     -FX     root             __         111.55 es Sun Mar 4 20:14
mpd4     -       root             ttyp0        0.08 es Sun Mar 4 20:13
mpd4     -       root             ttyp0      393.50 es Sun Mar 4 20:12
mpd4     -F      root             __           0.03 es Sun Mar 4 20:04
mpd4     -       root             ttyp0        0.08 es Sun Mar 4 20:04
mpd4     -F      root             __           0.03 es Sun Mar 4 20:03
mpd4     -       root             ttyp0        0.08 es Sun Mar 4 20:03
mpd4     -       root             ttyp0      104.34 es Sun Mar 4 20:02