Fedora 28 Won't Power Off

Have you checked if any ACPI/APIC options are disabled in the bios?

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown in
the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some of
the various proposed remedies, as they apply to my setup. I am using an LVM
partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-13 11:35:36 UTC

I'd like to make a general comment that "since forever" any kind of
power related command and <especially> shutdowns issued using Desktop
events have always been notoriously varied and unreliable. There has
been and still is no consistent way to execute a shutdown if you
compare any 2 Desktops.

The reason is relatively simple...
Although "shutdown -h now" or its systemd equivalent "systemctl
shutdown" can be invoked, both of these will generally shutdown and
power off the machine in a very abrupt way without regard to hanging
tasks like slow unmounts, data transfers in progress and more. In
other words, you risk data corruption.
So, Desktops often want to delay shutdown but there is generally no
assured way to know when all important processes have stopped, so time
delays of x number of seconds are generally inserted before complete
shutdown. Those add up and the User may wonder why the system just
won't shutdown quickly.
And then, there is always the possibility that something is hanging
and preventing shutdown.

If your distro is now built on systemd (Fedora 28 is), then you can
retrieve the syslog of the previous boot, the following does that and
displays last entries first

journalctl -r -b -1

Or, the following displays your syslog entries in real time. If you
know your system won't shutdown anyway, you can at least watch what is
happening during your attempted shutdown

journalctl -f

So, some suggestions...
Try another Desktop.
If you <really> want to shut down and power off quickly and you <know>
you don't have any data corruption concerns, then run either "shutdown
-h now" or "systemclt shutdown" in a console.

Tony

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-13 12:09:01 UTC

This post might be inappropriate. Click to display it.

DJA

2018-06-14 08:06:02 UTC

Post by Tony Su
Forgot to include in my previous post...
Another primary reason which makes shutdowns problematic is a legacy
concern from SysVinit and although partially addressed in systemd, not
yet completely addressed... The problem of spawning child processes
which can become orphaned.
systemd make progress by in principle invoking and managing child
processes within the same Control Group as the parent process so that
shutting down the parent process also shuts down all child processes
within that Control Group, but it doesn't seem to be perfect.
As long as orphaned processes that don't shutdown on their own can
still exist, shutdowns will always be problematic.
As I described, since this was a "standard feature" of SysVinit, there
is plenty of code published that invokes processes in a way that keeps
them "managed" so can be shutdown, but you have to go out of your way
to do this. By default, anything invoked in a SysVinit system is
unmanaged, with no likely way to track shutdown dependencies.
Tony

[snip]

Executive Summary: Someone broke something and no knows what, let alone
how to fix it. :D

Post by Tony Su

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.

[snip]

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

David Brown

2018-06-13 13:26:57 UTC

Post by Tony Su
I'd like to make a general comment that "since forever" any kind of
power related command and <especially> shutdowns issued using Desktop
events have always been notoriously varied and unreliable. There has
been and still is no consistent way to execute a shutdown if you
compare any 2 Desktops.
The reason is relatively simple...
Although "shutdown -h now" or its systemd equivalent "systemctl
shutdown" can be invoked, both of these will generally shutdown and
power off the machine in a very abrupt way without regard to hanging
tasks like slow unmounts, data transfers in progress and more. In
other words, you risk data corruption.

Not to be too impolite, but this is nonsense. Shutdown before systemd
is reasonably robust, at least scripts will generally be run, and init
will wait until all of the shutdown scripts have been run.

Then, it will invoke 'sync()' a few times with a delay to handle the
case where something wasn't able to be unmounted, and then invoke the
'reboot(2)' system call, telling the kernel to either shutdown or
reboot.

With systemd, it is pretty much the same, but systemd does a bit
better of a job of tracking what is happening. Regardless, it still
will invoke the 'reboot(2)' system call.

Reboot does some cleanup in the kernel, and then, for the case where a
shutdown was requrested (instead of a reboot), calls
"machine_power_off()". For x86, this generally uses ACPI to request
state G2, which physically powers off the machine.

Historically, (as in > 15 years ago), ACPI wasn't always reliable, and
some distros disabled it. I have no idea why a modern distribution
would do this, but I have also been seeing it on some modern machines.

ACPI shutdown should be reliable on any computer you are likely to be
running. If Fedora 28 isn't powering off, and you see a log message
about it reaching machine shutdown, I would look into why ACPI
shutdown isn't working in the kernel. It isn't likely to be something
from userspace, or at least if it is, there should be messages (maybe
press escape to see the messages).

David

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-14 08:14:06 UTC

Not to be too impolite, but this is nonsense. Shutdown before systemd
is reasonably robust, at least scripts will generally be run, and init
will wait until all of the shutdown scripts have been run.

That reflects my experience since about 1997. IOW, it worked.

Then, it will invoke 'sync()' a few times with a delay to handle the
case where something wasn't able to be unmounted, and then invoke the
'reboot(2)' system call, telling the kernel to either shutdown or
reboot.
With systemd, it is pretty much the same, but systemd does a bit
better of a job of tracking what is happening. Regardless, it still
will invoke the 'reboot(2)' system call.
Reboot does some cleanup in the kernel, and then, for the case where a
shutdown was requrested (instead of a reboot), calls
"machine_power_off()". For x86, this generally uses ACPI to request
state G2, which physically powers off the machine.
Historically, (as in > 15 years ago), ACPI wasn't always reliable, and
some distros disabled it. I have no idea why a modern distribution
would do this, but I have also been seeing it on some modern machines.

I haven't had to to this since the Redhat and Fedora Core days.

ACPI shutdown should be reliable on any computer you are likely to be
running. If Fedora 28 isn't powering off, and you see a log message
about it reaching machine shutdown, I would look into why ACPI
shutdown isn't working in the kernel.

How do I do this and what do I look for?

It isn't likely to be something
from userspace, or at least if it is, there should be messages (maybe
press escape to see the messages).
David

I can paste in some of the shutdown log from journalctl if that would
help, but as you know, there's a lot there. I don't know if journalctl
(which I hate) can push the log to a text file or another box.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-14 08:02:33 UTC

Maybe that's true. But in my experience in using desktops since oh,
Redhat started doing desktops, my systems have rarely had problems
shutting down. Even when there was the occasional problem, there was
always a fix. Doing substantial googling shows me that this current
problem is neither rare nor clearly understood.

Post by Tony Su
The reason is relatively simple...
Although "shutdown -h now" or its systemd equivalent "systemctl
shutdown" can be invoked, both of these will generally shutdown and
power off the machine in a very abrupt way without regard to hanging
tasks like slow unmounts, data transfers in progress and more. In
other words, you risk data corruption.
So, Desktops often want to delay shutdown but there is generally no
assured way to know when all important processes have stopped, so time
delays of x number of seconds are generally inserted before complete
shutdown. Those add up and the User may wonder why the system just
won't shutdown quickly.
And then, there is always the possibility that something is hanging
and preventing shutdown.

Without disputing the explanation (cuz you see I'm not qualified), I can
only say "Lame". How long has Linux been around, with desktops (I'm
using KDE by the way)?

Post by Tony Su
If your distro is now built on systemd (Fedora 28 is), then you can
retrieve the syslog of the previous boot, the following does that and
displays last entries first
journalctl -r -b -1

The problem is that the journal gets shut down before the system
completes the shutdown process. I can watch it by escaping from the
desktop. In fact I've removed rhgb and quiet from grub2.

Post by Tony Su
Or, the following displays your syslog entries in real time. If you
know your system won't shutdown anyway, you can at least watch what is
happening during your attempted shutdown
journalctl -f

I will try that. Although the shutdown process is waaay long and convoluted.

Post by Tony Su
So, some suggestions...
Try another Desktop.

Not a solution, but a way of ignoring the problem. Not much different
from "Linux has a problem? Try another OS*". ;)

Post by Tony Su
If you <really> want to shut down and power off quickly and you <know>
you don't have any data corruption concerns, then run either "shutdown
-h now" or "systemclt shutdown" in a console.
Tony

I don't need a quick shutdown. I need a shutdown.

* Actually, since systemd, Linux seems like another OS. Certainly not as
robust as what I grew up on. But that's another story.

Post by Tony Su

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-15 07:08:34 UTC

On 06/13/2018 04:35 AM, Tony Su wrote:

[snip]

This turned out to be of no use since any console from which this gets
run is one of the first processes/apps to get killed on shutdown (poweroff).

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-14 07:48:48 UTC

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

This is an older MB. There are no relevant ACPI options in the BIOS.
This problem has only started on recent Fedora versions. Intermittently
since maybe 27?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-25 21:18:04 UTC

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

For completeness. I had no ACPI/APIC options disabled in the BIOS. In
fact, there are essentially no such options. Just enabled/disabled for
ACPI. It's an old MB.

When I disabled ACPI, Linux refused to complete booting, complaining
about lack of ACPI support in the BIOS. I supposed I'd have to also
disable/remove acipd to deal with that problem (and no doubt cause more
user irritations).

Also, I am using the Nouveau driver for an older Nvidia card. I cannot
use a native driver because apparently, it needs to be patched to work
with newer iterations of Linux | systemd. I don't really feel like
mucking about with that unless I have to.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Gus Wirth

2018-06-14 21:15:13 UTC

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown
in the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some
of the various proposed remedies, as they apply to my setup. I am using
an LVM partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA

I experienced a similar problem when trying to shutdown from single mode
after doing some system maintenance. It turns out there was a user space
program, /usr/sbin/alsactl, that refused to shut down even with a kill
-9 directed at it. I don't even know how that is possible.

As an experiment, you can try to minimize the running processes before
shutting down. Start with getting rid of the GUI environment. Shut down
all your GUI applications. At a command prompt do the following:

$ sudo telinit 3

That should get rid of your GUI and drop you to a virtual console in
multiuser mode waiting for your login. That is also supposed to get rid
of all user processes except it doesn't, at least with KDE.

Log in as root.

Check to see what is running and kill everything that is not root or system.

# ps -ef

That will list all the running processes.

I sometimes just check to see what files are open on /home :

# lsof /home

and then I kill those processes.

# kill <process_id>

You should now be at a fairly minimal running state. Try shutting down:

# shutdown -h now

or

# halt -p

If that allows the system to shut down normally then the problem is in a
user process somewhere. If not, then it may be a kernel issue.

If you have another computer and both computers have serial ports, you
might consider setting up a serial console and turning on debugging to
help see what is going on.

Gus

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-16 06:17:58 UTC

Post by Gus Wirth

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.

[snip]

Post by Gus Wirth

Post by DJA
Any suggestions? TIA

I experienced a similar problem when trying to shutdown from single mode
after doing some system maintenance. It turns out there was a user space
program, /usr/sbin/alsactl, that refused to shut down even with a kill
-9 directed at it. I don't even know how that is possible.
As an experiment, you can try to minimize the running processes before
shutting down. Start with getting rid of the GUI environment. Shut down
$ sudo telinit 3
That should get rid of your GUI and drop you to a virtual console in
multiuser mode waiting for your login. That is also supposed to get rid
of all user processes except it doesn't, at least with KDE.
Log in as root.
Check to see what is running and kill everything that is not root or system.
# ps -ef
That will list all the running processes.
# lsof /home
and then I kill those processes.
# kill <process_id>
# shutdown -h now
or
# halt -p
If that allows the system to shut down normally then the problem is in a
user process somewhere. If not, then it may be a kernel issue.

Thanks for the tips.

I tried everything you suggested above. I killed every process that
didn't look to be kernel related (firewall, NetworkManager, DHCP, etc.
Even dbus) I did get a power off on halt one time, but have not been
able to repeat that since.

Post by Gus Wirth
If you have another computer and both computers have serial ports, you
might consider setting up a serial console and turning on debugging to
help see what is going on.
Gus

I don't have any serial ports for this.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-14 05:53:07 UTC

Try booting into emergency mode (not rescue – rescue is the equivalent of the old runlevel 1, and emergency mode is even more basic). Then try the command “poweroff -f -f”. See if that works to power off the system. BTW, don’t let the system catch fire

Shutdown -h now will do a clean shutdown and clean up all the systemd services. Poweroff will do the same thing, but if you specify -f twice, it should bypass all clean-shutdown.

This is the most basic test you can do. If the system doesn’t power off this way, it’s something kernel or ACPI/APIC related. If the system does power off this way, then you should be looking at something in systemd, a service, or maybe a socket or mount, preventing a full shutdown.

Sent from Mail for Windows 10

From: Dante Lanznaster <mailto:***@gmail.com>
Sent: Wednesday, June 13, 2018 12:07 AM
To: Main Discussion List for KPLUG <mailto:kplug-***@kernel-panic.org>
Subject: Re: Fedora 28 Won't Power Off

Have you checked if any ACPI/APIC options are disabled in the bios?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Gus Wirth

2018-06-15 16:15:08 UTC

Post by Kevin Keane Subscription
Try booting into emergency mode (not rescue – rescue is the
equivalent of the old runlevel 1, and emergency mode is even more
basic). Then try the command “poweroff -f -f”. See if that works to
power off the system. BTW, don’t let the system catch fire

How do you boot into emergency mode? I haven't heard of that before. Is
that something that systemd does?

Gus

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-16 06:36:43 UTC

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown
in the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some
of the various proposed remedies, as they apply to my setup. I am using
an LVM partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA

I have noticed some things which may be normal, but are curious:

During the shutdown process I see references to systemd stopping fsck on
some partitions. I don't have fsck running explicitly on anything as far
as I know.

As added info, I mount a directory from my fileserver with NFSv4 to
/var/nfsmounts/sevenhome. Entry in /etc/fstab:

seven:/home /var/nfsmounts/sevenhome nfs4
rw,_netdev,intr,retry=10,timeo=15,bg,sync 0 0

The fileserver (seven) is a VM running under KVM.

Unmounting /sevenhome before shutdown has no effect on the shutdown
process. Although I have seen messages saying unmounting /var failed.

All my partitions except /boot were created using LVM. I have two
physical drives, with all partitions created with LVM except /boot.

--- Physical volume ---
PV Name /dev/sdb2
VG Name proteus0
PV Size 728.00 GiB / not usable 4.00 MiB
Allocatable yes
PE Size 4.00 MiB
Total PE 186368
Free PE 1
Allocated PE 186367
PV UUID HiDZvD-gVEL-QNYZ-KDez-LZaH-kzOQ-zBCzso

--- Physical volume ---
PV Name /dev/sda2
VG Name vg_proteus
PV Size <595.92 GiB / not usable 1.81 MiB
Allocatable yes (but full)
PE Size 4.00 MiB
Total PE 152555
Free PE 0
Allocated PE 152555
PV UUID 3tH1YM-dWrN-jLZH-7jQj-KiNo-UKdf-XU78iE

/dev/sdb2 is a new drive which replaced /dev/sda2. Volume groups and
logical volumes are duplicated on /dev/sdb2 although of course with
different names. I do not mount any of the old partitions from sda2 on
sdb2. (sda1 and sdb1 are /boot partitons on their respective drives.)

I will gladly provide additional info as needed (and according to my skilz).

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-17 00:27:31 UTC

Emergency mode is simply another systemd target. So in the Grub screen, you type e to edit the boot configuration, then scroll down to all the kernel arguments, and you add the word “emergency” at the end.

The other targets are single (or rescue), corresponding to the old runlevel 1, multi-user (I believe) for the old runlevel 3, or graphical is the equivalent of the old runlevel 5.

Emergency mode is almost the same as booting off a CD into rescue mode, sans the CD. The root partition will be mounted read-only, and basically nothing will be started.

Sent from Mail for Windows 10

From: Gus Wirth <mailto:***@gmail.com>
Sent: Friday, June 15, 2018 9:15 AM
To: kplug-***@kernel-panic.org <mailto:kplug-***@kernel-panic.org>
Subject: Re: Fedora 28 Won't Power Off

How do you boot into emergency mode? I haven't heard of that before. Is
that something that systemd does?

Gus

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Gus Wirth

2018-06-17 02:34:51 UTC

Post by Kevin Keane Subscription
Emergency mode is simply another systemd target. So in the Grub screen,
you type e to edit the boot configuration, then scroll down to all the
kernel arguments, and you add the word “emergency” at the end.
The other targets are single (or rescue), corresponding to the old
runlevel 1, multi-user (I believe) for the old runlevel 3, or graphical
is the equivalent of the old runlevel 5.
Emergency mode is almost the same as booting off a CD into rescue mode,
sans the CD. The root partition will be mounted read-only, and basically
nothing will be started.

OK. Found that in the man page along with a bunch of other stuff that
systemd can do.

$ man systemd.special

This sounds like something DJA can try to get the absolute minimum
system running and then shutting down with the "poweroff -f -f" . If
that can work consistently then slowly start adding stuff back in, as in
"poweroff -f" (only one -f).

If the computer can't be shut down consistently from emergency mode with
a -f -f then I would say it's a kernel problem.

If it is a kernel problem then the solution will be to either wait for a
kernel update to fix it or to roll back to a kernel that worked. To do
that you'll need to download the kernels and associated packages from

https://koji.fedoraproject.org/koji/

Do a search for kernel. Look for an older kernel that supports your
system. For Fedora 28 I found this:

https://koji.fedoraproject.org/koji/buildinfo?buildID=1021748

It's a 4.15.0 kernel. This is a kernel line that was used in Fedora 26
and Fedora 27.

Find out what kernel packages are installed on your system:

$ rpm -qa "kernel*"

Find all the ones that are unique. For me, I can look in /boot and see
that I have a 4.16.15 kernel installed along with some others so I do this:

$ rpm -qa "kernel*" |grep 4.16.15
kernel-modules-4.16.15-300.fc28.x86_64
kernel-headers-4.16.15-300.fc28.x86_64
kernel-core-4.16.15-300.fc28.x86_64
kernel-modules-extra-4.16.15-300.fc28.x86_64
kernel-debug-devel-4.16.15-300.fc28.x86_64
kernel-4.16.15-300.fc28.x86_64
kernel-devel-4.16.15-300.fc28.x86_64

You will need to download and install those packages. The only thing
that will be tricky is the kernel-headers package. For some reason
Fedora allows only one of those packages so you'll have to uninstall it
before installing the older one. You'll have to use "rpm --nodeps -e
kernel-headers" because of a lot of stupid dependencies. You don't
really need it unless you're building against the kernel, which might
happen if you are running something like VirtualBox.

It's possible to use kernels from Fedora 27 in Fedora 28, so if that
doesn't help you can try a different kernel version.

If the older kernel works, you can prevent the system from upgrading the
kernel by putting an entry in /etc/dnf/dnfconf

excludepkgs=kernel*

Gus

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-18 19:45:50 UTC

Boy,
You guys are rough on me...
But that's OK :)

I looked into shutdown/poweroff problems a short time before and as
systemd was first implemented,
And I can guarantee you that I never saw anyone describe a sure-fire
way to shut down a system that was recalcitrant.

Granted,
If all that was running on a system was simple, well-behaving stuff,
then it's unlikely any shutdown problems would be enoucntered.

But,
If you were running something that typically had long running
processes and were not manageable by SysVinit, then no solution would
be possible and that problem was widely discussed.

Main culprits I ran into were
Enterprise Virtualization - Couldn't find a way for a HostOS to <know>
that all Guests were shut down so that it could in turn shut down
without special scripting. And, that's still the case today, AFAIK
HostOS shutdowns have to be specially handled, no "out of the box"
automatic solution is provided.

Database application - Generally no way for the HostOS to know if a
data transfer has completed (only the Database App knows this). This
always has to be addressed to avoid data corruption.

Infrequently, but not non-existent of course is unmounting. Sometimes
mounts just don't unmount on their own and they <shouldn't> if a data
transfer is in progress.

The point is that a poweroff/shutdown today is still a mostly "dumb"
procedure without insight into whether apps are ready to terminate,
and even the apps themselves are sometimes not smart enough to know
themselves.

And,
I'd suggest that the relics (no, still relevant) shutdown scripts that
abound for these known scenarios is evidence that problems exist else
those scripts wouldn't exist in the first place.

Tony

Post by Gus Wirth

Post by Kevin Keane Subscription
Emergency mode is simply another systemd target. So in the Grub screen,
you type e to edit the boot configuration, then scroll down to all the
kernel arguments, and you add the word “emergency” at the end.
The other targets are single (or rescue), corresponding to the old
runlevel 1, multi-user (I believe) for the old runlevel 3, or graphical is
the equivalent of the old runlevel 5.
Emergency mode is almost the same as booting off a CD into rescue mode,
sans the CD. The root partition will be mounted read-only, and basically
nothing will be started.

OK. Found that in the man page along with a bunch of other stuff that
systemd can do.
$ man systemd.special
This sounds like something DJA can try to get the absolute minimum system
running and then shutting down with the "poweroff -f -f" . If that can work
consistently then slowly start adding stuff back in, as in "poweroff -f"
(only one -f).
If the computer can't be shut down consistently from emergency mode with a
-f -f then I would say it's a kernel problem.
If it is a kernel problem then the solution will be to either wait for a
kernel update to fix it or to roll back to a kernel that worked. To do that
you'll need to download the kernels and associated packages from
https://koji.fedoraproject.org/koji/
Do a search for kernel. Look for an older kernel that supports your system.
https://koji.fedoraproject.org/koji/buildinfo?buildID=1021748
It's a 4.15.0 kernel. This is a kernel line that was used in Fedora 26 and
Fedora 27.
$ rpm -qa "kernel*"
Find all the ones that are unique. For me, I can look in /boot and see that
$ rpm -qa "kernel*" |grep 4.16.15
kernel-modules-4.16.15-300.fc28.x86_64
kernel-headers-4.16.15-300.fc28.x86_64
kernel-core-4.16.15-300.fc28.x86_64
kernel-modules-extra-4.16.15-300.fc28.x86_64
kernel-debug-devel-4.16.15-300.fc28.x86_64
kernel-4.16.15-300.fc28.x86_64
kernel-devel-4.16.15-300.fc28.x86_64
You will need to download and install those packages. The only thing that
will be tricky is the kernel-headers package. For some reason Fedora allows
only one of those packages so you'll have to uninstall it before installing
the older one. You'll have to use "rpm --nodeps -e kernel-headers" because
of a lot of stupid dependencies. You don't really need it unless you're
building against the kernel, which might happen if you are running something
like VirtualBox.
It's possible to use kernels from Fedora 27 in Fedora 28, so if that doesn't
help you can try a different kernel version.
If the older kernel works, you can prevent the system from upgrading the
kernel by putting an entry in /etc/dnf/dnfconf
excludepkgs=kernel*
Gus
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-19 06:35:43 UTC

Post by Tony Su
Boy,
You guys are rough on me...
But that's OK :)
I looked into shutdown/poweroff problems a short time before and as
systemd was first implemented,
And I can guarantee you that I never saw anyone describe a sure-fire
way to shut down a system that was recalcitrant.

[snip]

Post by Tony Su
The point is that a poweroff/shutdown today is still a mostly "dumb"
procedure without insight into whether apps are ready to terminate,
and even the apps themselves are sometimes not smart enough to know
themselves.
And,
I'd suggest that the relics (no, still relevant) shutdown scripts that
abound for these known scenarios is evidence that problems exist else
those scripts wouldn't exist in the first place.
Tony

Well then. I've been thinking about rebuilding my Win7 game box which
died about 4 years ago. I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's
time to give up the fight and take a serious look at Mac or Win10. I
know they shut down properly.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

David Brown

2018-06-19 14:48:50 UTC

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

Interesting. My experience has been pretty much the opposite.
Systemd has made my machines more reliable than they were before. I
run Gentoo, which is highly configurable, and I specifically select
the systemd configuration instead of init scripts because of that.

As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

Admittedly, it is a little difficult to tell precisely what distros
are being used, but according to distrowatch.com, the top ten
distributions are, in order:

- Linux Mint (based on Ubuntu, based on Debian)
- Ubuntu (based on Debian)
- Debian
- Mangeia (based on Red Hat)
- Fedora (based on Red Hat)
- SUSE (I wouldn't describe this as based on Red Hat, more just that
it uses the same package manager)
- Arch (It's own distro, not based on anything, pacman package
manager)
- CentOS (based on Red Hat)
- PCLinuxOS (Not sure what to call this, RPM packages, with APT)
- Slackware (Based on SLS, pkgtool as the "package manager" if you
could call it that)

In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu. I ran Fedora for a while,
and periodically ran into problems building things, like Android, or
Zephyr because dependencies weren't quite what was expected. I now
run Gentoo, and still occasionally have to setup an Ubuntu docker
image to be able to make everything work.

Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

If I were going to make a blanket statement, it would be something
like: "Enterprise tends to use something based on Red Hat. Open
source development tends to use Ubuntu. Individual users tend to use
either Ubuntu, or Mint, which is a fork of it."

There is definitely a multiple-world phenomemon, though, between
enterprise Linux use and other users. Fedora seems to be the one
distro that crosses over that boundary, but its use is still fairly
uncommon in the open source world.

David

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-19 15:58:44 UTC

I'd say that a major reason for Ubuntu's popularity is the Canonical
which literally poured resources into the whole Ubuntu ecosystem over
the past 12 years or so... But has been less active over the past year
instead beginning to prune technologies and projects that aren't as
popular.

That probably led to Ubuntu being the choice base OS and build system
for a great many software sectors, in my case I notice that
practically all security apps and especially pen testing are built
solely for APT.

But,
That's not to say that you run into difficulties with FOSS running on
RPM systems...
Some build services like the Open Build Service provide a way for
anyone with source to build a number of targets like package
management systems, distros, distro versions and more... With just a
few clicks instead of laboriously setting up build systems for each
target.

And,
Aside from those few specific software sectors, I hardly find problems
finding RPM software nowadays...
No major software I know of can't run on an RPM system that I know of.

Tony

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

Interesting. My experience has been pretty much the opposite.
Systemd has made my machines more reliable than they were before. I
run Gentoo, which is highly configurable, and I specifically select
the systemd configuration instead of init scripts because of that.
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.
Admittedly, it is a little difficult to tell precisely what distros
are being used, but according to distrowatch.com, the top ten
- Linux Mint (based on Ubuntu, based on Debian)
- Ubuntu (based on Debian)
- Debian
- Mangeia (based on Red Hat)
- Fedora (based on Red Hat)
- SUSE (I wouldn't describe this as based on Red Hat, more just that
it uses the same package manager)
- Arch (It's own distro, not based on anything, pacman package
manager)
- CentOS (based on Red Hat)
- PCLinuxOS (Not sure what to call this, RPM packages, with APT)
- Slackware (Based on SLS, pkgtool as the "package manager" if you
could call it that)
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu. I ran Fedora for a while,
and periodically ran into problems building things, like Android, or
Zephyr because dependencies weren't quite what was expected. I now
run Gentoo, and still occasionally have to setup an Ubuntu docker
image to be able to make everything work.
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).
If I were going to make a blanket statement, it would be something
like: "Enterprise tends to use something based on Red Hat. Open
source development tends to use Ubuntu. Individual users tend to use
either Ubuntu, or Mint, which is a fork of it."
There is definitely a multiple-world phenomemon, though, between
enterprise Linux use and other users. Fedora seems to be the one
distro that crosses over that boundary, but its use is still fairly
uncommon in the open source world.
David
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tracy Reed

2018-06-19 19:27:36 UTC

Post by David Brown
Admittedly, it is a little difficult to tell precisely what distros
are being used, but according to distrowatch.com, the top ten
- Linux Mint (based on Ubuntu, based on Debian)
- Ubuntu (based on Debian)
- Debian
- Mangeia (based on Red Hat)
- Fedora (based on Red Hat)
- SUSE (I wouldn't describe this as based on Red Hat, more just that
it uses the same package manager)
- Arch (It's own distro, not based on anything, pacman package
manager)
- CentOS (based on Red Hat)
- PCLinuxOS (Not sure what to call this, RPM packages, with APT)
- Slackware (Based on SLS, pkgtool as the "package manager" if you
could call it that)

How on earth are they calculating this? Mangeia (never heard of it!)
more popular than CentOS and Fedora?

Why is Mint so popular? I was never really interested in experimenting
with different distros and haven't tried anything but
RHEL/CentOS/Fedora/Ubuntu in ages. I've heard a lot about Arch lately
too, mostly memes about Arch users always having to point out they use
Arch. Although no Arch user has never pointed out to me that they use
Arch because I've never met one as far as I know.

Post by David Brown
There is definitely a multiple-world phenomemon, though, between
enterprise Linux use and other users. Fedora seems to be the one
distro that crosses over that boundary, but its use is still fairly
uncommon in the open source world.

I don't care for Fedora because the lifecycle is too short. I don't want
to have to reinstall my box every 6 months.

--
Tracy Reed
http://tracyreed.org
Digital signature attached for your safety.

David Brown

2018-06-19 21:24:03 UTC

This post might be inappropriate. Click to display it.

Tracy Reed

2018-06-19 23:25:05 UTC

Post by David Brown
I've never worked with anyone that has used a Red Hat/Centos/SUSE
(rpm) based distribution other than Fedora. I hear from people that
Enterprise users like to use Red Hat, but I've never worked with one
of these.

I run CentOS because it's what most of the production systems I deal
with run. And I like that patches are available for a long time so I
don't have to reinstall often.

Post by David Brown
I think the big issue with my "world" is that these are people that
have Linux machines to do software development on. These are not
machines that generally are running some kind of application (meaning
something like a web-facing app, or database, or something that would
have many users, but not users that are directly logged into the
machine).

Yeah...I do a tiny bit of software development (python, small programs
and web app stuff) and want my system to mirror prod as much as
possible.

I really don't understand the obsession with distributions aside from
the installer (which you only deal with once per machine) and the
package manager. I don't care what the desktop looks like etc. I spend
time looking at my apps and data, not the superficial eye candy.

Post by David Brown
Oh, I guess I should also say that I'm an Amazon Linux user (which is
RPM based), since that is kind of the default image on EC2. Virtual
machines where I care about doing non-trivial tasks, though, will
usually be installed as Ubuntu.

Isn't Amazon Linux RHEL/CentOS based?

--
Tracy Reed
http://tracyreed.org
Digital signature attached for your safety.

Frojon

2018-06-20 05:28:02 UTC

I keep using CentOS for monitoring systems and webservers and it has
served well.

Last production system that was not CentOS was SUSE for Groupwise... It
sure was a lot easier than my current Exchange Servers.

Zenoss has run now for around four years on a CentOS 1U from up at
Penguin Computers.

Going to replace Zenoss soon, but forget right now what my research
picked out and it will likely be on CentOS.

On personal systems - normally dual boot Fedora and Win 10 these days.
Mess with Ubuntu and Mint at times, but I like Fedora.

Frojon Banwell

Post by Tracy Reed

I run CentOS because it's what most of the production systems I deal
with run. And I like that patches are available for a long time so I
don't have to reinstall often.

Yeah...I do a tiny bit of software development (python, small programs
and web app stuff) and want my system to mirror prod as much as
possible.
I really don't understand the obsession with distributions aside from
the installer (which you only deal with once per machine) and the
package manager. I don't care what the desktop looks like etc. I spend
time looking at my apps and data, not the superficial eye candy.

Isn't Amazon Linux RHEL/CentOS based?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Rich Ernst

2018-06-19 23:12:46 UTC

I think Mint became very popular when Ubuntu went to their Unity(?)
desktop. I hated it, and found Mint had desktop more familiar and
similar to Windows 7. Since it's also Ubuntu based, I switched back
then. I recommend it for folks who are interested in trying Linux
because of that familiarity.

Rich

Post by Tracy Reed
How on earth are they calculating this? Mangeia (never heard of it!)
more popular than CentOS and Fedora?
Why is Mint so popular?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Robert Donovan

2018-06-25 14:50:41 UTC

I actually go back an forth between Korora(fedora based) Mint, and
Manjaro(arch based) I tried Arch, but found its bleeding edge, rolling
release approach to be a bit time consuming to deal with in a daily-use OS.
Manjaro uses the most recent stable versions of all the arch packages,
which avoids most of the upstream problems getting passed along. The Arch
community was pretty helpful when I had questions and I like the system,
but I need to get work done and didn't have time to solve a puzzle every
third update or so. YMMV.

Post by Tracy Reed

How on earth are they calculating this? Mangeia (never heard of it!)
more popular than CentOS and Fedora?
Why is Mint so popular? I was never really interested in experimenting
with different distros and haven't tried anything but
RHEL/CentOS/Fedora/Ubuntu in ages. I've heard a lot about Arch lately
too, mostly memes about Arch users always having to point out they use
Arch. Although no Arch user has never pointed out to me that they use
Arch because I've never met one as far as I know.

I don't care for Fedora because the lifecycle is too short. I don't want
to have to reinstall my box every 6 months.
--
Tracy Reed
http://tracyreed.org
Digital signature attached for your safety.
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-20 06:21:00 UTC

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

I don't want to have to tweak things to get them to work. I just want to
install the apps I need - out of the box - and do work.

Post by Kevin Keane Subscription
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

[snip of list of list of distros, most of which I've been aware of for
years.]

How many of these distros now use systemd? Does Redhat not pretty much
control systemd development and functionality? If that is indeed true,
are not nearly all apps dependent on systemd, and therefore Redhat?
That's what I meant.

A few years ago, I was actually strongly considering moving to Debian -
until it became clear that that distro would also be moving to systemd.

Post by Kevin Keane Subscription
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu.

I can't speak to what the Linux world thinks am using. I don't know
every person who uses Linux. Admittedly, I haven't researched usage
statistics because, well it's irrelevant to me: I use what I like and am
comfortable with. For the past 20 years or so, I've used Redhat --> Fedora.

Post by Kevin Keane Subscription
I ran Fedora for a while,
and periodically ran into problems building things, like Android, or
Zephyr because dependencies weren't quite what was expected. I now
run Gentoo, and still occasionally have to setup an Ubuntu docker
image to be able to make everything work.

I don't build things. I'm not a developer. So, interesting, but again
irrelevant to my own computing needs.

Post by Kevin Keane Subscription
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

See above.

Post by Kevin Keane Subscription
If I were going to make a blanket statement, it would be something
like: "Enterprise tends to use something based on Red Hat. Open
source development tends to use Ubuntu. Individual users tend to use
either Ubuntu, or Mint, which is a fork of it."

I'm not really interested in what I should tend to use. Right now I use
Fedora 28 (workstations), Scientific Linux 6 (VM host), and Ipfire
(router/firewall).

Post by Kevin Keane Subscription
There is definitely a multiple-world phenomemon, though, between
enterprise Linux use and other users. Fedora seems to be the one
distro that crosses over that boundary, but its use is still fairly
uncommon in the open source world.
David

So, how do I get my box to shutdown like it used to?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-20 05:50:47 UTC

I wouldn’t say that Fedora is “based on” RedHat, although they really are in the same family. If anything, RedHat is based on Fedora (but that’s not entirely accurate, either). Fedora is basically the beta version, the sandbox, for what RedHat will do a year or more later.

Usually, RedHat will take a particular Fedora version – usually every third one or so, but lately they seem to have been waiting longer – and uses it as a starting point for the next release. I’m anxious to find out what RedHat 8 will bring when it eventually comes out.

Meanwhile, CentOS is an almost direct clone of RedHat. There are a few subtle differences. I use CentOS for most of my own systems, and almost exclusively RedHat at work. Surprisingly, RedHat is significantly more stable than CentOS. If you have the budget for it, I would highly recommend using the actual RedHat for production systems, instead of CentOS. Of course, I’m spoiled. We have a campus license for RedHat.

Sent from Mail for Windows 10

From: David Brown <mailto:***@davidb.org>
Sent: Tuesday, June 19, 2018 7:49 AM
To: Main Discussion List for KPLUG <mailto:kplug-***@kernel-panic.org>
Subject: Re: Fedora 28 Won't Power Off

I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

Interesting. My experience has been pretty much the opposite.
Systemd has made my machines more reliable than they were before. I
run Gentoo, which is highly configurable, and I specifically select
the systemd configuration instead of init scripts because of that.

As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

Admittedly, it is a little difficult to tell precisely what distros
are being used, but according to distrowatch.com, the top ten
distributions are, in order:

- Linux Mint (based on Ubuntu, based on Debian)
- Ubuntu (based on Debian)
- Debian
- Mangeia (based on Red Hat)
- Fedora (based on Red Hat)
- SUSE (I wouldn't describe this as based on Red Hat, more just that
    it uses the same package manager)
- Arch (It's own distro, not based on anything, pacman package
    manager)
- CentOS (based on Red Hat)
- PCLinuxOS (Not sure what to call this, RPM packages, with APT)
- Slackware (Based on SLS, pkgtool as the "package manager" if you
    could call it that)

In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu. I ran Fedora for a while,
and periodically ran into problems building things, like Android, or
Zephyr because dependencies weren't quite what was expected. I now
run Gentoo, and still occasionally have to setup an Ubuntu docker
image to be able to make everything work.

Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

If I were going to make a blanket statement, it would be something
like: "Enterprise tends to use something based on Red Hat. Open
source development tends to use Ubuntu. Individual users tend to use
either Ubuntu, or Mint, which is a fork of it."

There is definitely a multiple-world phenomemon, though, between
enterprise Linux use and other users. Fedora seems to be the one
distro that crosses over that boundary, but its use is still fairly
uncommon in the open source world.

David

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-20 06:42:45 UTC

System is a RedHat project, that much is true. But that doesn’t really mean that “all distros are RedHat” just as all distros using the same kernel doesn’t mean they are all the same. Personally, when it comes to systemd vs initscripts, I used to be agnostic. The more I work with systemd, the more I like it, though.

Among the mainstream distros, there are basically two main lines: the RedHat line, and the Debian line. Most, but not all, mainstream distros are derived from or related to one of those two.

And let’s not forget that there are many very important non-mainstream distros. Many of those rely on busybox rather than bash and other tools. Alpine comes to mind. OpenWRT/DD-WRT (which, with a heavy heart, I highly advise against).

BTW, one more interesting, and somewhat important, distro is Raspbian. It’s in the Debian family, of course. There also is a Fedora-derived distro for the Raspberrry if you prefer that route.

Sent from Mail for Windows 10

From: DJA <mailto:***@codermotor.com>
Sent: Tuesday, June 19, 2018 11:22 PM
To: kplug-***@kernel-panic.org <mailto:kplug-***@kernel-panic.org>
Subject: Re: Fedora 28 Won't Power Off

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

I don't want to have to tweak things to get them to work. I just want to
install the apps I need - out of the box - and do work.

Post by Kevin Keane Subscription
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

Post by Kevin Keane Subscription
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu.

I don't build things. I'm not a developer. So, interesting, but again
irrelevant to my own computing needs.

Post by Kevin Keane Subscription
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

See above.

I'm not really interested in what I should tend to use. Right now I use
Fedora 28 (workstations), Scientific Linux 6 (VM host), and Ipfire
(router/firewall).

So, how do I get my box to shutdown like it used to?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Brad Beyenhof

2018-06-20 18:02:57 UTC

Post by Kevin Keane Subscription
And let’s not forget that there are many very important non-mainstream distros. Many of those rely on busybox rather than bash and other tools. Alpine comes to mind. OpenWRT/DD-WRT (which, with a heavy heart, I highly advise against).

--
Brad Beyenhof . . . . . . . . . . . . . . . . http://augmentedfourth.com
There are only two kinds of men: the righteous, who believe themselves
sinners; the rest, sinners, who believe themselves righteous.
~ Blaise Pascal (1623–1662)
--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-20 19:39:04 UTC

Post by Brad Beyenhof

I've installed OpenWRT with the Luci frontend in my and a friend's
access points and am happy with it.

I use Ipfire for my router and firewall. https://www.ipfire.org/ Very
well maintained and kept up to date, with security being the main focus.
I have never had a single problem.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-21 10:24:12 UTC

IMO
If you want to shutdown your system with certainty, and because you're
certain that you no longer have any essential tasks still running,
Just avoid using the Desktop shutdown/poweroff if it's giving you trouble.

In a console,
You can still execute as always

shutdown -h now

Slightly less overbearing, you can now also run the following commands
according to the following MAN page
https://www.freedesktop.org/software/systemd/man/systemd-halt.service.html

systemctl halt
systemctl poweroff
systemctl reboot
systemctl kexec

Note also that if you're shutting down a number of machines, these
commands should work against numerous systems regardless of distro,
distro version or possible Desktop if they all run systemd.

Tony

On Tue, Jun 19, 2018 at 11:42 PM, Kevin Keane Subscription

Post by Kevin Keane Subscription
System is a RedHat project, that much is true. But that doesn’t really mean that “all distros are RedHat” just as all distros using the same kernel doesn’t mean they are all the same. Personally, when it comes to systemd vs initscripts, I used to be agnostic. The more I work with systemd, the more I like it, though.
Among the mainstream distros, there are basically two main lines: the RedHat line, and the Debian line. Most, but not all, mainstream distros are derived from or related to one of those two.
And let’s not forget that there are many very important non-mainstream distros. Many of those rely on busybox rather than bash and other tools. Alpine comes to mind. OpenWRT/DD-WRT (which, with a heavy heart, I highly advise against).
BTW, one more interesting, and somewhat important, distro is Raspbian. It’s in the Debian family, of course. There also is a Fedora-derived distro for the Raspberrry if you prefer that route.
Sent from Mail for Windows 10
Sent: Tuesday, June 19, 2018 11:22 PM
Subject: Re: Fedora 28 Won't Power Off

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

I don't want to have to tweak things to get them to work. I just want to
install the apps I need - out of the box - and do work.

Post by David Brown
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

[snip of list of list of distros, most of which I've been aware of for
years.]
How many of these distros now use systemd? Does Redhat not pretty much
control systemd development and functionality? If that is indeed true,
are not nearly all apps dependent on systemd, and therefore Redhat?
That's what I meant.
A few years ago, I was actually strongly considering moving to Debian -
until it became clear that that distro would also be moving to systemd.

Post by David Brown
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu.

Post by David Brown
I ran Fedora for a while,
and periodically ran into problems building things, like Android, or
Zephyr because dependencies weren't quite what was expected. I now
run Gentoo, and still occasionally have to setup an Ubuntu docker
image to be able to make everything work.

I don't build things. I'm not a developer. So, interesting, but again
irrelevant to my own computing needs.

Post by David Brown
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

See above.

Post by David Brown
If I were going to make a blanket statement, it would be something
like: "Enterprise tends to use something based on Red Hat. Open
source development tends to use Ubuntu. Individual users tend to use
either Ubuntu, or Mint, which is a fork of it."

I'm not really interested in what I should tend to use. Right now I use
Fedora 28 (workstations), Scientific Linux 6 (VM host), and Ipfire
(router/firewall).

So, how do I get my box to shutdown like it used to?
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-21 10:38:44 UTC

systemd is largely developed in Fedora.
If you look at the historical roots of systemd, it was primarily built
by two individuals (I can't remember off the top of my head besides
Poeterring), one employed at Fedora and the other at openSUSE. After
some initial releases, the openSUSE employee shifted over to Fedora as
well.

Typically any new systemd releases are introduced in Fedora, then
within maybe 3 weeks or so become available also in openSUSE
Tumbleweed (The openSUSE rolling release which generally gets
everything new first). As the "enterprise" based version, openSUSE
LEAP won't see the new version until bugs are thoroughly worked out in
Tumbleweed.

As for RPi,
There are a number of distros that release versions for RPi, including
openSUSE. Raspian is probably is the most official distro you'll find
recommended on the RPi website but you can look for others. For
openSUSE, this provides a nice way for Users to experience and use the
same tools they are familiar with on x64 on to a RPI. The following
page for example has everything you'd need to get started installing
numerous images of both LEAP and Tumbleweed with various Desktops on
the RPI3 (There are other pages for the original RPI/Zero and RPI2)

https://en.opensuse.org/HCL:Raspberry_Pi3

Tony

Post by Tony Su
IMO
If you want to shutdown your system with certainty, and because you're
certain that you no longer have any essential tasks still running,
Just avoid using the Desktop shutdown/poweroff if it's giving you trouble.
In a console,
You can still execute as always
shutdown -h now
Slightly less overbearing, you can now also run the following commands
according to the following MAN page
https://www.freedesktop.org/software/systemd/man/systemd-halt.service.html
systemctl halt
systemctl poweroff
systemctl reboot
systemctl kexec
Note also that if you're shutting down a number of machines, these
commands should work against numerous systems regardless of distro,
distro version or possible Desktop if they all run systemd.
Tony
On Tue, Jun 19, 2018 at 11:42 PM, Kevin Keane Subscription

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

I don't want to have to tweak things to get them to work. I just want to
install the apps I need - out of the box - and do work.

Post by David Brown
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

[snip of list of list of distros, most of which I've been aware of for
years.]
How many of these distros now use systemd? Does Redhat not pretty much
control systemd development and functionality? If that is indeed true,
are not nearly all apps dependent on systemd, and therefore Redhat?
That's what I meant.
A few years ago, I was actually strongly considering moving to Debian -
until it became clear that that distro would also be moving to systemd.

Post by David Brown
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu.

I don't build things. I'm not a developer. So, interesting, but again
irrelevant to my own computing needs.

Post by David Brown
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

See above.

I'm not really interested in what I should tend to use. Right now I use
Fedora 28 (workstations), Scientific Linux 6 (VM host), and Ipfire
(router/firewall).

So, how do I get my box to shutdown like it used to?
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-21 10:50:45 UTC

systemd is a RHEL project only in the sense that Fedora is owned by,
and supported by RHEL.
I haven't tracked recent RHEL policy regarding systemd, but for a very
long time RHEL resisted implementing systemd, at one point an
announcement was made of its intention to migrate to systemd, but then
was withdrawn.

In any case,
People who have used or know systemd
Know that typically for at least a year when a systemd object is
introduced that replaces an init object, the init commands will still
be supported. Eventually though, that 100% compatibility won't exist
and then it becomes YMMV... You might have to use the systemd
commands.

And, systemd isn't replacing SysVinit in one massive overnight move.
systemd supports anything that still runs only as an init, and that
includes many applications.
In those cases, you'll still have the init way of executing (because
that's the real structure) but systemd will also dynamically create
the means to use "systemctl" to manage and execute the init as well,
even without a Unit file (which is the systemd configuration file for
each app, service, and similar level process).

So, people who prefer init should <usually> have a not so painful way
of migrating to systemd, but eventually must when 100% compatibility
is dropped (the stick) or to take advantage of features and
versatility (the carrot).

Tony

Post by Tony Su
systemd is largely developed in Fedora.
If you look at the historical roots of systemd, it was primarily built
by two individuals (I can't remember off the top of my head besides
Poeterring), one employed at Fedora and the other at openSUSE. After
some initial releases, the openSUSE employee shifted over to Fedora as
well.
Typically any new systemd releases are introduced in Fedora, then
within maybe 3 weeks or so become available also in openSUSE
Tumbleweed (The openSUSE rolling release which generally gets
everything new first). As the "enterprise" based version, openSUSE
LEAP won't see the new version until bugs are thoroughly worked out in
Tumbleweed.
As for RPi,
There are a number of distros that release versions for RPi, including
openSUSE. Raspian is probably is the most official distro you'll find
recommended on the RPi website but you can look for others. For
openSUSE, this provides a nice way for Users to experience and use the
same tools they are familiar with on x64 on to a RPI. The following
page for example has everything you'd need to get started installing
numerous images of both LEAP and Tumbleweed with various Desktops on
the RPI3 (There are other pages for the original RPI/Zero and RPI2)
https://en.opensuse.org/HCL:Raspberry_Pi3
Tony

Post by DJA
I only do art now anyway. With Linux, in my
experience, having gotten less reliable since systemd came to town, and
as all distros seem to now be essentially Redhat variants, maybe it's

I don't want to have to tweak things to get them to work. I just want to
install the apps I need - out of the box - and do work.

Post by David Brown
As far as "all distros" being essentially Redhat variants, I'm not
sure where you get that.

[snip of list of list of distros, most of which I've been aware of for
years.]
How many of these distros now use systemd? Does Redhat not pretty much
control systemd development and functionality? If that is indeed true,
are not nearly all apps dependent on systemd, and therefore Redhat?
That's what I meant.
A few years ago, I was actually strongly considering moving to Debian -
until it became clear that that distro would also be moving to systemd.

Post by David Brown
In my experience, in the open source world development world, it is
assumed that you will be running Ubuntu.

I don't build things. I'm not a developer. So, interesting, but again
irrelevant to my own computing needs.

Post by David Brown
Building Android from source explicitly requires either Mac OS, or
Ubuntu Linux. Other distributions are not supported (but aren't
usually that hard to make work).

See above.

I'm not really interested in what I should tend to use. Right now I use
Fedora 28 (workstations), Scientific Linux 6 (VM host), and Ipfire
(router/firewall).

So, how do I get my box to shutdown like it used to?
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-22 06:05:32 UTC

Executed from a KDE Konsole:
Does not work. Says "powering off", then synchronized and stop both
disks. Backlight and all functionality go off on my keyboard. Machine
does not power off.

Post by Tony Su
Slightly less overbearing, you can now also run the following commands
according to the following MAN page
https://www.freedesktop.org/software/systemd/man/systemd-halt.service.html

The following were done by switching to the <F2> console from the GUI
login screen. IOW, I did not log in to KDE.

Post by Tony Su
systemctl halt

Same results as above.

Post by Tony Su
systemctl poweroff

Ditto.

Post by Tony Su
systemctl reboot

Again.

Post by Tony Su
systemctl kexec

"Can't find ESP partition mount point" is echoed to screen. And prompt
returns. Doing any of the above commands results in same shutdown
messages without an actual hardware power off.

Post by Tony Su
Note also that if you're shutting down a number of machines, these
commands should work against numerous systems regardless of distro,
distro version or possible Desktop if they all run systemd.
Tony

Thanks for the suggestions.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-22 12:24:15 UTC

Don't know what to say,
I've personally never had a situation where "shutdown -h now" or
"systemctl halt" didn't power off.
Those are about as possible to shutdown and power off as you can get.

Maybe if you do an Internet search on why <those> aren't working for
you might turn up something...

Tony

Does not work. Says "powering off", then synchronized and stop both disks.
Backlight and all functionality go off on my keyboard. Machine does not
power off.

Post by Tony Su
Slightly less overbearing, you can now also run the following commands
according to the following MAN page
https://www.freedesktop.org/software/systemd/man/systemd-halt.service.html

The following were done by switching to the <F2> console from the GUI login
screen. IOW, I did not log in to KDE.

Post by Tony Su
systemctl halt

Same results as above.

Post by Tony Su
systemctl poweroff

Ditto.

Post by Tony Su
systemctl reboot

Again.

Post by Tony Su
systemctl kexec

"Can't find ESP partition mount point" is echoed to screen. And prompt
returns. Doing any of the above commands results in same shutdown messages
without an actual hardware power off.

Thanks for the suggestions.
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

David Brown

2018-06-23 01:34:49 UTC

Post by Tony Su
Don't know what to say,
I've personally never had a situation where "shutdown -h now" or
"systemctl halt" didn't power off.
Those are about as possible to shutdown and power off as you can get.

If you want to bypass all of init/systemd, and just ask the kernel to
turn off the computer (as an experiment), do something like:

# sync
wait a few seconds
# poweroff -f -f

(Note, use 'poweroff' or 'hat', not the "shutdown" command)
which should just call the kernel's reboot call asking it to turn off
the computer. If this doesn't turn it off, it is a kernel/ACPI issue.
Otherwise, it has something to do with the systemd system.

David

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-23 06:33:37 UTC

If you want to bypass all of init/systemd, and just ask the kernel to
# sync
wait a few seconds
# poweroff -f -f
(Note, use 'poweroff' or 'hat', not the "shutdown" command)
which should just call the kernel's reboot call asking it to turn off
the computer. If this doesn't turn it off, it is a kernel/ACPI issue.
Otherwise, it has something to do with the systemd system.
David

Tried this. Same results. The last messages say "Powering off" followed
by "KVM: exiting hardware virtualization" then syncing SCSI caches and
stopping each drive.

If I shutdown from KDE, I get the

"kernel not configured for semaphores (System V IPC). Not using udev
synchronization code" messages.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Rich Ernst

2018-06-23 15:52:57 UTC

Are you running any virtual machines?

If not, do you have cpu/hardware virtualization feature turned off in the BIOS?

Why this would SUDDENLY make a difference, I've no idea, but it was a thought.

Rich

Tried this. Same results. The last messages say "Powering off" followed by
"KVM: exiting hardware virtualization" then syncing SCSI caches and stopping
each drive.
If I shutdown from KDE, I get the
"kernel not configured for semaphores (System V IPC). Not using udev
synchronization code" messages.
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-25 21:05:37 UTC

Post by Rich Ernst
Are you running any virtual machines?
If not, do you have cpu/hardware virtualization feature turned off in the BIOS?

I was not. I turned off the BIOS support. The only difference at
shutdown (without scouring logs) was the elimination of one message
about KVM shutting down.

Post by Rich Ernst
Why this would SUDDENLY make a difference, I've no idea, but it was a thought.
Rich

Obviously, me either. But thanks for the suggestion.

Post by Rich Ernst

Tried this. Same results. The last messages say "Powering off" followed by
"KVM: exiting hardware virtualization" then syncing SCSI caches and stopping
each drive.
If I shutdown from KDE, I get the
"kernel not configured for semaphores (System V IPC). Not using udev
synchronization code" messages.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-21 06:56:29 UTC

To be clear: OpenWRT and DD-WRT are valiant efforts, and the guys who make them really deserve props for accomplishing something amazing. And I also have to qualify my statement: it depends on what you are trying to accomplish. As always in life, just because a particular tool performed poorly for me doesn’t mean it’s a bad tool, period.

The use cases where OpenWRT an DD-WRT shine are situations where you need a fairly large network for a short term on a shoestring budget. Say you need to cover a volunteer-run music festival or church gathering in a large park. You will spend a *lot* more effort than with commercial offerings, but spend a lot fewer $$$.

Also, if you need a lab environment and want to experiment with network topologies and routing protocols, these distros may be great.

If it’s your main router that connects you to the Internet – beware.

These projects are far too big for one person to run. It can often take months before a new router model is supported, and then there often aren’t any updates for years. I had put DD-WRT on my ASUS RT-12N router a couple of years ago (I actually wanted OpenWRT, but it wasn’t out for this router until a year or so later). Once I put DD-WRT on the router, it worked reasonably well. But there was no update for a whopping 7 years. Eventually, I had to turn off https because the DD-WRT version was so old that it only supported SSLv2, which most browsers nowadays block.

The next problem is that when updates do appear, you can’t just flash it on your router. You have to write down all your settings, do a factory reset, then install the new version, and configure it from scratch. Think you could do a backup of your config, and restore it on the new version? Sorry, doesn’t work. Backups can only be restored to the same version of DD-WRT that created it.

Basically, once you install it, you are stuck with that version unless you can afford several days of downtime. Upgrading means either taking down your Internet connection for a day or so (and keep in mind, you can’t Google anything without the Internet), or replacing the router hardware.

And the third problem is documentation, or lack thereof. Theoretically, there is a database on the Web site with all the supported router models, and the corresponding version of DD-WRT that you need, but in reality that hasn’t been updated in probably a decade. In reality, you have to visit the support forum, find the thread that discusses your particular router model, and wade through 30+ pages of posts to figure it out.

My experience is with DD-WRT. OpenWRT is slightly better, but does have many of the same problems.

That said: once I got my router going, it worked reasonably well for 10 years or so. It was always flaky, and we had to power-cycle it roughly once per week. Whether that was a DD-WRT bug or a hardware issue, I will never know. I suspect a hardware issue, actually.

Today, I’m using an Engadget WiFi bridge for the same purpose, and it is both more reliable, and much easier to set up. And much neater, too, because it uses PoE.

Sent from Mail for Windows 10

From: Brad Beyenhof <mailto:***@icloud.com>
Sent: Wednesday, June 20, 2018 11:04 AM
To: KPLUG Mailman <mailto:kplug-***@kernel-panic.org>
Subject: Router Linux distros (was Re: Fedora 28 Won't Power Off)

Post by Kevin Keane Subscription

And let’s not forget that there are many very important non-mainstream distros. Many of those rely on busybox rather than bash and other tools. Alpine comes to mind. OpenWRT/DD-WRT (which, with a heavy heart, I highly advise against).

You advise against both OpenWRT *and* DD-WRT, or just DD-WRT? Why is that?

I'm about to move, and was thinking about using OpenWRT again in the new home's network. I've been using the standard Linksys firmware lately, but I did buy a router that could handle Linux.

--
Brad Beyenhof . . . . . . . . . . . . . . . . http://augmentedfourth.com
There are only two kinds of men: the righteous, who believe themselves
sinners; the rest, sinners, who believe themselves righteous.
~ Blaise Pascal (1623–1662)

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Brad Beyenhof

2018-06-21 16:43:09 UTC

[snip]

Thanks for the detailed analysis. I'll keep those factors in mind.

Post by Kevin Keane Subscription
My experience is with DD-WRT. OpenWRT is slightly better, but does have many of the same problems.

I've run OpenWRT with good success. My favorite part was the package manager for software customization, where DD-WRT just has different sizes of binary blob so you have to pick your full package list before you start. The last time I set up a home network, I was trying to use signal repeaters, which relied on using the vendor firmware. The signal repeaters didn't work as expected, but after that failure I didn't want to take down my in-use home network just to rebuild it on OpenWRT.

Post by Kevin Keane Subscription
That said: once I got my router going, it worked reasonably well for 10 years or so. It was always flaky, and we had to power-cycle it roughly once per week. Whether that was a DD-WRT bug or a hardware issue, I will never know. I suspect a hardware issue, actually.

I had DD-WRT on a WRT54G that ran fine for several years, but somehow every time I wanted to log in to change anything (once every few months), the UI was locked up and I needed to re-boot. I didn't have any complaints with the actual routing, though.

I also used Gargoyle once to set up a spare WRT54G as a wireless *client* so I could connect a few network devices without wireless cards (notably, a laser printer) some distance from the main modem & router. I think I originally configured that one at an InstallFest Gus Wirth hosted at his house, actually.
--
Brad Beyenhof . . . . . . . . . . . . . . . . http://augmentedfourth.com
Every man takes the limits of his own field of vision for the limits
of the world.
~ Arthur Schopenhauer, German philosopher (1788-1860)

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-22 03:55:24 UTC

Yes, I also like the OpenWRT package manager. DD-WRT can use it as well, actually, but it requires more trickery. I suppose most of my issues happened because I thought of the device as a *router* rather than as a Linux computer, with the corresponding different expectations. I only chose DD-WRT because OpenWRT didn’t support my device yet at the time. And I replaced the stock firmware because I needed RIP, and a wireless client.

And on repeaters: yeah, I hear you. Don’t do that. You need one integrated system, not a hodgepodge of different networks, which is what these early repeaters (and many SOHO devices even today) will give you.

Until maybe two years ago, all the solutions were either abysmal, or enterprise-grade expensive. Sometimes both. Today, I wouldn’t think twice – just use Ubiquiti.

Sent from Mail for Windows 10

From: Brad Beyenhof <mailto:***@icloud.com>
Sent: Thursday, June 21, 2018 9:43 AM
To: Kevin Keane Subscription <mailto:***@kkeane.com> ; KPLUG Mailman <mailto:kplug-***@kernel-panic.org>
Subject: Re: Router Linux distros (was Re: Fedora 28 Won't Power Off)

[snip]

Thanks for the detailed analysis. I'll keep those factors in mind.

Post by Kevin Keane Subscription
My experience is with DD-WRT. OpenWRT is slightly better, but does have many of the same problems.

I had DD-WRT on a WRT54G that ran fine for several years, but somehow every time I wanted to log in to change anything (once every few months), the UI was locked up and I needed to re-boot. I didn't have any complaints with the actual routing, though.

I also used Gargoyle once to set up a spare WRT54G as a wireless *client* so I could connect a few network devices without wireless cards (notably, a laser printer) some distance from the main modem & router. I think I originally configured that one at an InstallFest Gus Wirth hosted at his house, actually.

--
Brad Beyenhof . . . . . . . . . . . . . . . . http://augmentedfourth.com
Every man takes the limits of his own field of vision for the limits
of the world.
~ Arthur Schopenhauer, German philosopher (1788-1860)

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-06-26 03:45:34 UTC

On the Nvidia driver: it actually comes as an installer that works very well. It does some compiling.

A few things to note about that driver:

* You have to reinstall after every kernel update. Since I maintain one of our labs with quite a few computers, that was a fairly big undertaking. I created a cron script that, once an hour, checked if the driver was loaded, and if not, automatically recompiles it.
* Turn off secure boot. UEFI booting works fine, and there is a way to sign the driver when compiling it, but managing the key is a fairly big headache. If you don’t sign it and boot with secure boot, the driver will refuse to load.
* Installing with the installer is a snap. Uninstalling, on the other hand, is, to put it mildly, less than trivial. It just took me four days to recover from removing the Nvidia card from my main work system. I didn’t even get a login prompt on tty1 in multi-user mode (runlevel 3). And if you are unfortunate enough that graphics do show up, your keyboard won’t work any more. Even more infuriating: if you manually run startx, you’ll get graphics just fine! Here is what you have to do to recover:
* Switch to tty2 if you are still in text mode. If not, SSH into your system. The system runs fine, it’s just the console that is broken.
* Switch to multiuser mode.
* Find all the places where the nouveau driver is blacklisted.
* Rebuild the initramrd.
* Reinstall xorg-\* and mesa-\*
* I also had to putz around with Plymouth, Dracut, reinstalled all the kernel driver modules, played around with /etc/default/grub and rebuilt the grub config files, and more. Not sure if all of those steps are necessary.

Bottom line: if you need he features of the Nvidia driver, by all means install it. If not, stick with nouveau.

Sent from Mail for Windows 10

From: DJA <mailto:***@codermotor.com>
Sent: Monday, June 25, 2018 2:19 PM
To: kplug-***@kernel-panic.org <mailto:kplug-***@kernel-panic.org>
Subject: Re: Fedora 28 Won't Power Off

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Tony Su

2018-06-26 03:53:59 UTC

Look into whether your nVidia driver can be installed with DKMS
support to enable auto self-updating whenever your kernel is upgraded.

Tony

On Mon, Jun 25, 2018 at 8:45 PM, Kevin Keane Subscription

Post by Kevin Keane Subscription
On the Nvidia driver: it actually comes as an installer that works very well. It does some compiling.
* You have to reinstall after every kernel update. Since I maintain one of our labs with quite a few computers, that was a fairly big undertaking. I created a cron script that, once an hour, checked if the driver was loaded, and if not, automatically recompiles it.
* Turn off secure boot. UEFI booting works fine, and there is a way to sign the driver when compiling it, but managing the key is a fairly big headache. If you don’t sign it and boot with secure boot, the driver will refuse to load.
* Switch to tty2 if you are still in text mode. If not, SSH into your system. The system runs fine, it’s just the console that is broken.
* Switch to multiuser mode.
* Find all the places where the nouveau driver is blacklisted.
* Rebuild the initramrd.
* Reinstall xorg-\* and mesa-\*
* I also had to putz around with Plymouth, Dracut, reinstalled all the kernel driver modules, played around with /etc/default/grub and rebuilt the grub config files, and more. Not sure if all of those steps are necessary.
Bottom line: if you need he features of the Nvidia driver, by all means install it. If not, stick with nouveau.
Sent from Mail for Windows 10
Sent: Monday, June 25, 2018 2:19 PM
Subject: Re: Fedora 28 Won't Power Off

Post by Dante Lanznaster
Have you checked if any ACPI/APIC options are disabled in the bios?

For completeness. I had no ACPI/APIC options disabled in the BIOS. In
fact, there are essentially no such options. Just enabled/disabled for
ACPI. It's an old MB.
When I disabled ACPI, Linux refused to complete booting, complaining
about lack of ACPI support in the BIOS. I supposed I'd have to also
disable/remove acipd to deal with that problem (and no doubt cause more
user irritations).
Also, I am using the Nouveau driver for an older Nvidia card. I cannot
use a native driver because apparently, it needs to be patched to work
with newer iterations of Linux | systemd. I don't really feel like
mucking about with that unless I have to.

--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list
--
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-06-26 21:55:16 UTC

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown
in the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some
of the various proposed remedies, as they apply to my setup. I am using
an LVM partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA

From I can gleen from a lot off googling, this problem has been around
since at least Fedora 25, but is not confined to Fedora alone. It seems
to be related to a combination of SELinux and LVM. And is not
consistently repeatable.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-11-20 20:45:18 UTC

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown
in the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some
of the various proposed remedies, as they apply to my setup. I am using
an LVM partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA

I finally got this box to actually power off.

I upgraded to F29. That had the same problem. It seems Red Hat's
Bugzilla provides lots of complaints, but no solutions which actually
work for most people.

I had a couple of newer video cards sitting around, all Nvidia-based. I
installed one. Shutdown worked fine with the Nouveau driver.
Unfortunately, that card was had overheating problems.

I installed a newer card - GeForce 780ti. Power off no longer worked
with the Nouveau driver. Just for grins, I removed the Nouveau driver,
and installed NVidia's driver. Power off now works fine every time.

Problem solved*

So for me, apparently something related to the video driver was
preventing the PSU from powering off (although keyboard, mouse, USB and
drives all go down, leaving CPU and case fans running).

*I like some aspects of the Nouveau driver better: proper text video
resolution at boot up so I can see what's going on. Less time to
graphical log-in screen; not a PITA to deal with after each kernel
update (DKMS is supposed to fix that. We'll see).

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

Kevin Keane Subscription

2018-11-21 05:59:40 UTC

-----Original message-----

Sent: Tuesday, November 20, 2018 12:46 PM
Subject: Re: Fedora 28 Won't Power Off

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.
The console messages vary from shutdown to shutdown, and do not always
display the same final messages. But they do look similar to those shown
in the first post of the bug report below.
https://bugzilla.redhat.com/show_bug.cgi?id=1385432
That thread does not seem to show any solutions yet. I have tried some
of the various proposed remedies, as they apply to my setup. I am using
an LVM partitions.
I was getting the
"Kernel not configured for semaphores (System V IPC). Not using udev
synchronisation code."
error, but don't see it lately, since some kernel upgrades.
Any suggestions? TIA

I finally got this box to actually power off.
I upgraded to F29. That had the same problem. It seems Red Hat's
Bugzilla provides lots of complaints, but no solutions which actually
work for most people.
I had a couple of newer video cards sitting around, all Nvidia-based. I
installed one. Shutdown worked fine with the Nouveau driver.
Unfortunately, that card was had overheating problems.
I installed a newer card - GeForce 780ti. Power off no longer worked
with the Nouveau driver. Just for grins, I removed the Nouveau driver,
and installed NVidia's driver. Power off now works fine every time.
Problem solved*
So for me, apparently something related to the video driver was
preventing the PSU from powering off (although keyboard, mouse, USB and
drives all go down, leaving CPU and case fans running).
*I like some aspects of the Nouveau driver better: proper text video
resolution at boot up so I can see what's going on. Less time to
graphical log-in screen; not a PITA to deal with after each kernel
update (DKMS is supposed to fix that. We'll see).

Now that you narrowed it down, I may be able to contribute another piece of the puzzle. We had lots of systems with NVidia graphics cards, and RedHat 7. What I noticed was that whenever we restarted the systems, I'd get an abrt error. The restart completed without any further issues. That happened with the Nvidia driver; I think for us Nouveau was a little more stable (but transitioning back from NVidia to Nouveau is a major PITA). Interestingly, I have a similar problem with our new systems, which have Intel graphics cards.

My guess is that this crash happens during the very late stage of the shutdown.

There is one other problem with the NVidia driver: when installed normally, it is incompatible with Secure Boot because the driver is not signed. I'm not sure if Fedora supports Secure Boot in the first place, but it was an issue with RedHat.

As for the kernel update, since I had to manage it on a dozen identical computers, I had to automate things. My solution: I wrote a script that checks if the kernel modules exist. If not, it kicked off the Nvidia installer in unattended mode. I ran this script once an hour. Since all our software updates were installed during a middle-of-the-night maintenance window, the workstations would have the correct drivers by the morning.

The Nvidia driver offers another option: you can build it on one system (and provide a signing key), and then push the already-compiled version out to all your systems. If you add the key to your UEFI BIOS, you can even use the Nvidia driver with Secure Boot.

--
KPLUG-***@kernel-panic.org
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-list

DJA

2018-11-21 06:51:30 UTC

Post by Kevin Keane Subscription
-----Original message-----

Sent: Tuesday, November 20, 2018 12:46 PM

Post by DJA
I've been having a problem with Fedora 28 ever since an upgrade. The
system seems to go thru the shutdown process, but never powers off the
computer.

[snip]

I finally got this box to actually power off.

[snip]