10:26 a.m.
Running rawhide on my primary desktop for the last few years has always
been good for a surprise or two. It's part of the fun. But I'm
curious: is it just me, or is the experience a bit more rawer than usual
at the moment?
- nmh has some sort of locking trouble - commands just hang to the
point that only kill -9 will get rid of them. Local ext3 filesystem,
no NFS nastiness involved. Running the commands as root works. But
who wants to run MH as root?
- /tmp gets filled with tmprules-N.dat files. A bit of searching
points the finger at thunderbird on this one.
- X clients occasionally just exit, seemingly losing their connection
to the server. For whatever reason, panel applets (and the panel
itself) seem especially prone to this one.
- Sound doesn't work - all the files in /dev/snd are set to root-only
permissions. Looks like maybe the pulseaudio transition is stuck
partway?
So am I alone in all this?
Thanks,
jon
10:24 a.m.
On Mon, 2007-09-24 at 09:26 -0600, Jonathan Corbet wrote:
- X clients occasionally just exit, seemingly losing their
connection
to the server. For whatever reason, panel applets (and the panel
itself) seem especially prone to this one.
I haven't seen this. It might be good to attach strace to one and dump
the output somewhere to see what happens just before it exits.
- ajax
9:37 a.m.
New subject: An especially raw rawhide? (X client disconnects)
Adam Jackson <ajackson(a)redhat.com> wrote:
On Mon, 2007-09-24 at 09:26 -0600, Jonathan Corbet wrote:
> - X clients occasionally just exit, seemingly losing their connection
> to the server. For whatever reason, panel applets (and the panel
> itself) seem especially prone to this one.
I haven't seen this. It might be good to attach strace to one and dump
the output somewhere to see what happens just before it exits.
I've been trying. But, when you get one random client disconnecting
every hour or so, it's *really* hard to get a useful capture.
Emacs has died once or twice, printing a one-line "lost connection to X
server" message.
One interesting data point: I've dropped back to a 2.6.22 kernel and
have yet to see the problem. It's only been a couple of hours of work
time, so it's too soon to say, but there might be some sort of
unix-domain socket weirdness in 2.6.23-rc. If that is true, it would
sure be nice to get a handle on it before the final comes out.
Thanks,
jon
Jonathan Corbet / LWN.net / corbet(a)lwn.net
2:38 p.m.
New subject: An especially raw rawhide? (X disconnects)
Adam Jackson <ajackson(a)redhat.com> wrote:
On Mon, 2007-09-24 at 09:26 -0600, Jonathan Corbet wrote:
> - X clients occasionally just exit, seemingly losing their connection
> to the server. For whatever reason, panel applets (and the panel
> itself) seem especially prone to this one.
I haven't seen this. It might be good to attach strace to one and dump
the output somewhere to see what happens just before it exits.
One other data point: I just had another application fail with:
xv: xcb_xlib.c:41: xcb_xlib_lock: Assertion `!c->xlib.lock' failed.
Abort
This *might* be a completely different problem, though - if it happens,
it does so right at startup. The application starts correctly probably
14 times out of 15.
jon
2:47 p.m.
New subject: An especially raw rawhide? (X disconnects)
One other data point: I just had another application fail with:
xv: xcb_xlib.c:41: xcb_xlib_lock: Assertion `!c->xlib.lock'
failed.
Abort
This *might* be a completely different problem, though - if it happens,
it does so right at startup. The application starts correctly probably
14 times out of 15.
jon
I haven't really read this thread but, I tried boinc
(boinc.berkeley.edu) 64bit client few days ago and it failed everytime
with that same message.
4:07 a.m.
New subject: An especially raw rawhide? (X disconnects)
I am getting that same error (xv: xcb_xlib.c:41: xcb_xlib_lock:
Assertion `!c->xlib.lock' failed.) running some java apps.
A previous update fixed it but it is back now - not sure which is the
offending package but am sure someone else can point it out
On 9/26/07, Riku Seppälä <riku.seppala(a)kymp.net> wrote:
One other data point: I just had another application fail with:
> xv: xcb_xlib.c:41: xcb_xlib_lock: Assertion `!c->xlib.lock' failed.
> Abort
>
> This *might* be a completely different problem, though - if it happens,
> it does so right at startup. The application starts correctly probably
> 14 times out of 15.
>
> jon
>
>
I haven't really read this thread but, I tried boinc
(boinc.berkeley.edu) 64bit client few days ago and it failed everytime
with that same message.
--
fedora-test-list mailing list
fedora-test-list(a)redhat.com
To unsubscribe:
https://www.redhat.com/mailman/listinfo/fedora-test-list
10:44 a.m.
On Mon, 2007-09-24 at 09:26 -0600, Jonathan Corbet wrote:
- nmh has some sort of locking trouble - commands just hang to the
point that only kill -9 will get rid of them. Local ext3 filesystem,
no NFS nastiness involved. Running the commands as root works. But
who wants to run MH as root?
Might be worth running it under strace to get a better idea of what's
going on. I suspect you're the only person left in the world running
nmh ;-)
- Sound doesn't work - all the files in /dev/snd are set to
root-only
permissions. Looks like maybe the pulseaudio transition is stuck
partway?
They should have acls to allow your user access -- also worth seeing if
pulseaudio is actually running.
Jeremy
9:42 a.m.
New subject: An especially raw rawhide? (nmh and audio)
Jeremy Katz <katzj(a)redhat.com> wrote:
On Mon, 2007-09-24 at 09:26 -0600, Jonathan Corbet wrote:
> - nmh has some sort of locking trouble - commands just hang to the
> point that only kill -9 will get rid of them. Local ext3 filesystem,
> no NFS nastiness involved. Running the commands as root works. But
> who wants to run MH as root?
Might be worth running it under strace to get a better idea of what's
going on. I suspect you're the only person left in the world running
nmh ;-)
That's easy:
% strace scan
[...]
open("/h/corbet/Mail/context", O_RDWR|O_NONBLOCK) = 3
fcntl(3, F_SETLK, {type=F_WRLCK, whence=SEEK_SET, start=0, len=0}) = -1 ENOLCK (No locks
available)
The fcntl() call takes a while to time out, then the program just
loops and tries again.
> - Sound doesn't work - all the files in /dev/snd are set to
root-only
> permissions. Looks like maybe the pulseaudio transition is stuck
> partway?
They should have acls to allow your user access -- also worth seeing if
pulseaudio is actually running.
% getfacl /dev/snd/*
getfacl: Removing leading '/' from absolute path names
# file: dev/snd/controlC0
# owner: root
# group: root
user::rw-
group::rw-
other::rw-
...and it goes on in that vein forever. Who is supposed to be setting
those ACLs?
As for pulseaudio, I have:
3312 ? Sl 0:00 /usr/bin/pulseaudio --log-target=syslog
3599 ? S 0:00 /usr/libexec/pulse/gconf-helper
Thanks,
jon
Jonathan Corbet / LWN.net / corbet(a)lwn.net
10:04 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 08:42:18 -0600
lwn-ft(a)lwn.net (Jonathan Corbet) wrote:
% getfacl /dev/snd/*
getfacl: Removing leading '/' from absolute path names
# file: dev/snd/controlC0
# owner: root
# group: root
user::rw-
group::rw-
other::rw-
...and it goes on in that vein forever. Who is supposed to be setting
those ACLs?
I think this is an selinux policy issue we just found yesterday. If
you boot with 'enforcing=0' as a kernel argument does it get better?
This will put selinux into a permissive state from the very get go,
instead of relying upon the config file to adjust it at some point
during boot. I just had to do this to get pulseaudio functional for me
as well.
--
Jesse Keating
Fedora -- All my bits are free, are yours?
10:18 a.m.
New subject: An especially raw rawhide? (nmh and audio)
Jesse Keating wrote:
On Tue, 25 Sep 2007 08:42:18 -0600
lwn-ft(a)lwn.net (Jonathan Corbet) wrote:
> % getfacl /dev/snd/*
> getfacl: Removing leading '/' from absolute path names
> # file: dev/snd/controlC0
> # owner: root
> # group: root
> user::rw-
> group::rw-
> other::rw-
>
> ...and it goes on in that vein forever. Who is supposed to be setting
> those ACLs?
I think this is an selinux policy issue we just found yesterday. If
you boot with 'enforcing=0' as a kernel argument does it get better?
Do you mean selinux=0 here?
Rahul
10:34 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 20:48:53 +0530
Rahul Sundaram <sundaram(a)fedoraproject.org> wrote:
Do you mean selinux=0 here?
No, selinux=0 will completely turn it off and prevent any new labels
from being created, and cause a need for a mass relable next time you
want to turn it on. enforcing=0 will put selinux into permissive mode
where it will still log what would have been denied, but not actually
deny things. Extremely useful for debugging policy issues.
--
Jesse Keating
Fedora -- All my bits are free, are yours?
10:38 a.m.
New subject: An especially raw rawhide? (nmh and audio)
Jesse Keating <jkeating(a)redhat.com> wrote:
I think this is an selinux policy issue we just found yesterday. If
you boot with 'enforcing=0' as a kernel argument does it get better?
This will put selinux into a permissive state from the very get go,
instead of relying upon the config file to adjust it at some point
during boot.
I'm not running SELinux at all, so I don't think that would change
much...
Thanks,
jon
Jonathan Corbet / LWN.net / corbet(a)lwn.net
10:48 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 09:38:52 -0600
lwn-ft(a)lwn.net (Jonathan Corbet) wrote:
I'm not running SELinux at all, so I don't think that would
change
much...
How do you have it disabled though? I learned yesterday that there is
an important difference between having it disabled via boot time
argument (selinux=0 or enforcing=0) vs having it disabled in
the /etc/sysconfig/selinux config file (which is what Firstboot would
poke). If you have it disabled via the config file, that may be too
late for the hal startup to set the appropriate ACLs. It has to be
disabled at boot time, or you have to restart haldaemon after selinux
is shut off via the config file.
--
Jesse Keating
Fedora -- All my bits are free, are yours?
10:51 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 2007-09-25 at 11:48 -0400, Jesse Keating wrote:
On Tue, 25 Sep 2007 09:38:52 -0600
lwn-ft(a)lwn.net (Jonathan Corbet) wrote:
> I'm not running SELinux at all, so I don't think that would change
> much...
How do you have it disabled though? I learned yesterday that there is
an important difference between having it disabled via boot time
argument (selinux=0 or enforcing=0) vs having it disabled in
the /etc/sysconfig/selinux config file (which is what Firstboot would
poke).
Huh? If you have it disabled via /etc/selinux/config, then init doesn't
load a policy and pokes the kernel to disable the hooks. Which should
then be identical to selinux=0
Jeremy
11:03 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 11:51:33 -0400
Jeremy Katz <katzj(a)redhat.com> wrote:
Huh? If you have it disabled via /etc/selinux/config, then init
doesn't load a policy and pokes the kernel to disable the hooks.
Which should then be identical to selinux=0
Ah, maybe it's just a difference between booting with enforcing=0 and
having permissive be set in /etc/sysconfig/selinux.
--
Jesse Keating
Fedora -- All my bits are free, are yours?
11:29 a.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 2007-09-25 at 12:03 -0400, Jesse Keating wrote:
On Tue, 25 Sep 2007 11:51:33 -0400
Jeremy Katz <katzj(a)redhat.com> wrote:
> Huh? If you have it disabled via /etc/selinux/config, then init
> doesn't load a policy and pokes the kernel to disable the hooks.
> Which should then be identical to selinux=0
Ah, maybe it's just a difference between booting with enforcing=0 and
having permissive be set in /etc/sysconfig/selinux.
Shouldn't matter, as everything is allowed until policy is loaded.
--
Stephen Smalley
National Security Agency
1:37 p.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 12:29:11 -0400
Stephen Smalley <sds(a)tycho.nsa.gov> wrote:
> > Huh? If you have it disabled via /etc/selinux/config, then
init
> > doesn't load a policy and pokes the kernel to disable the hooks.
> > Which should then be identical to selinux=0
>
> Ah, maybe it's just a difference between booting with enforcing=0
> and having permissive be set in /etc/sysconfig/selinux.
Shouldn't matter, as everything is allowed until policy is loaded.
You guys are right, I was being led down the wrong path. Anyway what
I'm seeing, which is close to what Jonathan is seeing, may actually be
a ConsoleKit bug. https://bugzilla.redhat.com/show_bug.cgi?id=305801
Some debugging with Mr. Zeuthen led to this bug being filed.
--
Jesse Keating
Fedora -- All my bits are free, are yours?
1:48 p.m.
New subject: An especially raw rawhide? (nmh and audio)
On Tue, 25 Sep 2007 14:37:33 -0400
Jesse Keating <jkeating(a)redhat.com> wrote:
You guys are right, I was being led down the wrong path. Anyway
what
I'm seeing, which is close to what Jonathan is seeing, may actually be
a ConsoleKit bug. https://bugzilla.redhat.com/show_bug.cgi?id=305801
Some debugging with Mr. Zeuthen led to this bug being filed.
And he just came back and figured it out. thinkfinger installs a hal
info file which actually has some policy on it, but the policy isn't
correctly defined, so when hal goes to do all the ACL changes, it finds
an improper policy and bails out. Now David is going to fix it so that
hal doesn't bail on 3rd party stuff, and then reassign the bug to
thinkfinger to fix their policy (:
--
Jesse Keating
Fedora -- All my bits are free, are yours?
11:01 a.m.
On 09/24/2007 11:26 AM, Jonathan Corbet wrote:
- Sound doesn't work - all the files in /dev/snd are set to root-only
permissions. Looks like maybe the pulseaudio transition is stuck
partway?
That sounds familiar:
https://bugzilla.redhat.com/show_bug.cgi?id=287941
9:45 a.m.
New subject: An especially raw rawhide? (Sound)
Chuck Ebbert <cebbert(a)redhat.com> wrote:
On 09/24/2007 11:26 AM, Jonathan Corbet wrote:
>
> - Sound doesn't work - all the files in /dev/snd are set to root-only
> permissions. Looks like maybe the pulseaudio transition is stuck
> partway?
That sounds familiar:
https://bugzilla.redhat.com/show_bug.cgi?id=287941
Sort of...except that I'm running neither KDE nor SELinux.
Thanks,
jon
6078
days inactive
6080
days old
19 comments
9 participants
participants (9)
-
Adam Jackson -
Chuck Ebbert -
Dan Holwerda -
Jeremy Katz -
Jesse Keating -
Jonathan Corbet -
Rahul Sundaram -
Riku Seppälä -
Stephen Smalley