On Monday 23 January 2006 13:27, "John (J5) Palmieri" <johnp(a)redhat.com>
wrote:
On Sun, 2006-01-22 at 15:55 -0500, Dan Williams wrote:
> On Sun, 2006-01-22 at 15:12 +0000, Richard Hughes wrote:
> > On Sun, 2006-01-22 at 19:56 +0530, Rahul Sundaram wrote:
> > > Joachim Frieben wrote:
> > >
> > > >After udating "dbus" packages to version 0.60-6,
"avahi" and
> > > >"hal" daemons fail to start up correctly during system
boot.
> > > >This is very annoying when logging in to a "GNOME" session
> > > >as various applications such as the "gnome-power-manager"
stop
> > > >to work. Media detection is of course also out of order.
> > > >The breakage has appeared sometime between version 0.60-3 which
> > > >works correctly and the current version. Downgrading to version
> > > >0.60-3 allows to recover a working "GNOME" desktop.
> > > >
> > > >
> > > Looks like dbus is working fine but has a made a incompatible change and
> > > the other programs like avahi and g-p-m needs to be updated. Just a
guess.
> >
> > g-p-m is working fine with DBUS cvs -- but I believe J5 made some
> > changes to the DBUS package for rawhide to put stuff in different
> > directories. The breakage is probably due to that.
>
> That shouldn't "break stuff" unless avahi and/or hal use paths to
> dbus-send or stuff like that (dhclient-script does this, I've fixed it).
> Perhaps they do. Anyone care to check?
I'm putting a symlink in /usr/bin/ to "fix" this but applications should
fix this by looking in /bin. Also if you did not update your SELinux
policies you will get breakage. Other than that applications that don't
do anything wacky should not break.
--
John (J5) Palmieri <johnp(a)redhat.com>
The avahi startup issue appears to be caused by an issue with
selinux-policy-targeted-2.2.2-1 :
# audit2allow </var/log/audit/audit.log | grep avahi
allow avahi_t initrc_t:unix_stream_socket connectto;
# grep avahi-daemon /var/log/audit/audit.log
type=AVC msg=audit(1138039806.833:117): avc: denied { connectto } for pid=2696
comm="avahi-daemon" name="system_bus_socket"
scontext=system_u:system_r:avahi_t:s0 tcontext=system_u:system_r:initrc_t:s0
tclass=unix_stream_socket
This is weird, as the system-bus-socket seems to have the correct context:
$ ls -lZ /var/run/dbus/system_bus_socket
srwxrwxrwx root root system_u:object_r:system_dbusd_var_run_t
/var/run/dbus/system_bus_socket
Yet all dbus-daemon's FDs have context system_u:system_r:initrc_t :
# ls -Z /proc/$dbus_daemon_system_pid/fd
...system_u:system_r:initrc_t...
I think this is because /usr/bin/dbus-daemon, which HAD context
'system_u:object_r:system_dbusd_exec_t',
has now moved to /bin, where it now has context 'system_u:object_r:bin_t' .
I'm not sure that just creating links will rectify this ( /bin precedes /usr/bin in
the initscripts $PATH )-
selinux-policy needs updating to take account of the new dbus-daemon location.
Thanks & Regards,
Jason Vas Dias <jvdias(a)redhat.com>