I tried with IOMMU Enabled, and with IOMMU disabled and iommu=soft in
Grub - no improvement.
I suspect the problem lies in the graphical environment - the screen
froze for about 20 minutes but is now responding again. While frozen,
the system still served files via NFS to other systems. Certaily some of
the freezes seemed to occur in connection with mouse movements/button
presses.
Thought I might follow the wiki How to set up the graphical server with
Nouveau as an alternative to the nVidia driver. Now I get "Invalid magic
number" message, but there seems to be no posts that correct this. An
ideas please?
--
Grimble
Registered Linux User #450547
Machine 'Bach' running Plasma 5.20.4 on 5.10.43-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64

On 21/06/2021 15:41, Grimble wrote:
> I tried with IOMMU Enabled, and with IOMMU disabled and iommu=soft in
> Grub - no improvement.
> I suspect the problem lies in the graphical environment - the screen
> froze for about 20 minutes but is now responding again. While frozen,
> the system still served files via NFS to other systems. Certaily some of
> the freezes seemed to occur in connection with mouse movements/button
> presses.
> Thought I might follow the wiki How to set up the graphical server with
> Nouveau as an alternative to the nVidia driver. Now I get "Invalid magic
> number" message, but there seems to be no posts that correct this. An
> ideas please?
Seems like an update to kernel 5.10.43 failed part way through, so had
to drop back to 5.10.41. Change to nouveau worked but I'm not optimistic
the problem's solved - Chrome bombed as soon as I launched it.

--
Grimble
Registered Linux User #450547
Machine 'mozart' running Plasma 5.20.4 on 5.10.41-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64

On Tue, 22 Jun 2021 14:19:09 -0400, grimble <grimble@nomail.afraid.org> wrote:

> On 21/06/2021 15:41, Grimble wrote:
>> I tried with IOMMU Enabled, and with IOMMU disabled and iommu=soft in
>> Grub - no improvement.
>> I suspect the problem lies in the graphical environment - the screen
>> froze for about 20 minutes but is now responding again. While frozen,
>> the system still served files via NFS to other systems. Certaily some of
>> the freezes seemed to occur in connection with mouse movements/button
>> presses.
>> Thought I might follow the wiki How to set up the graphical server with
>> Nouveau as an alternative to the nVidia driver. Now I get "Invalid magic
>> number" message, but there seems to be no posts that correct this. An
>> ideas please?
> Seems like an update to kernel 5.10.43 failed part way through, so had
> to drop back to 5.10.41. Change to nouveau worked but I'm not optimistic
> the problem's solved - Chrome bombed as soon as I launched it.

More info about how the update failed is needed. Did it run out of space?

Regards, Dave Hodgins

--
Change dwhodgins@nomail.afraid.org to davidwhodgins@teksavvy.com for
email replies.

On 22/06/2021 22:13, David W. Hodgins wrote:
> On Tue, 22 Jun 2021 14:19:09 -0400, grimble <grimble@nomail.afraid.org>
> wrote:
>
>> On 21/06/2021 15:41, Grimble wrote:
>>> I tried with IOMMU Enabled, and with IOMMU disabled and iommu=soft in
>>> Grub - no improvement.
>>> I suspect the problem lies in the graphical environment - the screen
>>> froze for about 20 minutes but is now responding again. While frozen,
>>> the system still served files via NFS to other systems. Certaily some of
>>> the freezes seemed to occur in connection with mouse movements/button
>>> presses.
>>> Thought I might follow the wiki How to set up the graphical server with
>>> Nouveau as an alternative to the nVidia driver. Now I get "Invalid magic
>>> number" message, but there seems to be no posts that correct this. An
>>> ideas please?
>> Seems like an update to kernel 5.10.43 failed part way through, so had
>> to drop back to 5.10.41. Change to nouveau worked but I'm not optimistic
>> the problem's solved - Chrome bombed as soon as I launched it.
>
> More info about how the update failed is needed. Did it run out of space?
>
> Regards, Dave Hodgins
>
Very unlikely, 9+ GB available on /. I noticed several of the 5.10.43
files in /boot, such as vmlinuz- and sysvers- were 0 B.

--
Grimble
Registered Linux User #450547
Machine 'Bach' running Plasma 5.20.4 on 5.10.43-desktop-1.mga8 kernel.
Mageia release 8 (Official) for x86_64

On Wed, 23 Jun 2021 14:56:33 +0100, Grimble wrote:
> I noticed several of the 5.10.43
> files in /boot, such as vmlinuz- and sysvers- were 0 B.

Then those are failures. /boot should look something like this snippet:

239K Jun 11 02:33 config-5.10.43-desktop-1.mga8
4.0K Dec 26 08:34 dracut
4.0K Jun 14 09:21 grub2
17M Jun 14 09:12 initrd-5.10.43-desktop-1.mga8.img
33 Jun 14 09:13 initrd-desktop.img -> initrd-5.10.43-desktop-1.mga8.img
33 Jun 14 09:13 initrd.img -> initrd-5.10.43-desktop-1.mga8.img
193K Jun 11 02:33 symvers-5.10.43-desktop-1.mga8.xz
4.8M Jun 11 02:33 System.map-5.10.43-desktop-1.mga8
30 Jun 14 09:13 vmlinuz -> vmlinuz-5.10.43-desktop-1.mga8
7.7M Jun 11 02:33 vmlinuz-5.10.43-desktop-1.mga8
30 Jun 14 09:13 vmlinuz-desktop -> vmlinuz-5.10.43-desktop-1.mga8

On 23/06/2021 17:41, Bit Twister wrote:
> On Wed, 23 Jun 2021 14:56:33 +0100, Grimble wrote:
>> I noticed several of the 5.10.43
>> files in /boot, such as vmlinuz- and sysvers- were 0 B.
>
> Then those are failures. /boot should look something like this snippet:
>
> 239K Jun 11 02:33 config-5.10.43-desktop-1.mga8
> 4.0K Dec 26 08:34 dracut
> 4.0K Jun 14 09:21 grub2
> 17M Jun 14 09:12 initrd-5.10.43-desktop-1.mga8.img
> 33 Jun 14 09:13 initrd-desktop.img -> initrd-5.10.43-desktop-1.mga8.img
> 33 Jun 14 09:13 initrd.img -> initrd-5.10.43-desktop-1.mga8.img
> 193K Jun 11 02:33 symvers-5.10.43-desktop-1.mga8.xz
> 4.8M Jun 11 02:33 System.map-5.10.43-desktop-1.mga8
> 30 Jun 14 09:13 vmlinuz -> vmlinuz-5.10.43-desktop-1.mga8
> 7.7M Jun 11 02:33 vmlinuz-5.10.43-desktop-1.mga8
> 30 Jun 14 09:13 vmlinuz-desktop -> vmlinuz-5.10.43-desktop-1.mga8
>
Thanks, Bit. Fortunately 5.10.45 installed without problems. However,
we're still trying to find the root cause of this instability.
G

--
Grimble
Registered Linux User #450547
Machine 'Bach' running Plasma 5.20.4 on 5.10.45-desktop-2.mga8 kernel.
Mageia release 8 (Official) for x86_64

On Thu, 24 Jun 2021 12:55:34 +0100, Grimble wrote:
> On 23/06/2021 17:41, Bit Twister wrote:
>> On Wed, 23 Jun 2021 14:56:33 +0100, Grimble wrote:
>>> I noticed several of the 5.10.43
>>> files in /boot, such as vmlinuz- and sysvers- were 0 B.
>>
>> Then those are failures. /boot should look something like this snippet:
>>
>> 239K Jun 11 02:33 config-5.10.43-desktop-1.mga8
>> 4.0K Dec 26 08:34 dracut
>> 4.0K Jun 14 09:21 grub2
>> 17M Jun 14 09:12 initrd-5.10.43-desktop-1.mga8.img
>> 33 Jun 14 09:13 initrd-desktop.img -> initrd-5.10.43-desktop-1.mga8.img
>> 33 Jun 14 09:13 initrd.img -> initrd-5.10.43-desktop-1.mga8.img
>> 193K Jun 11 02:33 symvers-5.10.43-desktop-1.mga8.xz
>> 4.8M Jun 11 02:33 System.map-5.10.43-desktop-1.mga8
>> 30 Jun 14 09:13 vmlinuz -> vmlinuz-5.10.43-desktop-1.mga8
>> 7.7M Jun 11 02:33 vmlinuz-5.10.43-desktop-1.mga8
>> 30 Jun 14 09:13 vmlinuz-desktop -> vmlinuz-5.10.43-desktop-1.mga8
>>
> Thanks, Bit. Fortunately 5.10.45 installed without problems.

Well that is a step forward. Since the old kernels did not compile you
might consider using mcc to remove them.

> However,
> we're still trying to find the root cause of this instability.

Yes I have been following the thread. Maybe enabling core dump might give
you a chance to find out why.

$ cat enable_coredump.txt
#*********** start of enable_coredump.txt ****************************
# following are instructions for enabling core dumps.
# # "ulimit -c unlimited" has to executed during login either by
# each user in $HOME/.bash_profile or $HOME/.bashrc or globally
# for everyone. The following is globally.

# As root, paste the following in a root terminal:

echo '#!/bin/bash
# Enable core dump for each user
ulimit -c unlimited
#***** end of /etc/profile.d/xx_enable_coredump.sh
' > /etc/profile.d/xx_enable_coredump.sh

echo '#!/bin/csh
# Enable core dump for each user
ulimit -c unlimited
#***** end of /etc/profile.d/xx_enable_coredump.csh
' > /etc/profile.d/xx_enable_coredump.csh

chmod +x /etc/profile.d/xx_enable_coredump.sh
chmod +x /etc/profile.d/xx_enable_coredump.csh

#****************************************************************
# you need to make sysctl changes, as root, paste the following:
#****************************************************************

mkdir --parents /etc/sysctl.d/

echo '#**** start of /etc/sysctl.d/xx_enable_coredump.conf
# enable System Request debugging functionality of the kernel
kernel.sysrq = 1

# Enabling suid dump PID appending and set core location and name

kernel.core_uses_pid = 1
kernel.core_pattern = /var/tmp/%e_%p_%s.core
fs.suid_dumpable = 2

#********* end of /etc/sysctl.d/xx_enable_coredump.conf *********
' > /etc/sysctl.d/xx_enable_coredump.conf

#***********************************************
# reload all sysctl changes with the following:
#***********************************************

sysctl -a

#**************************************************
# add any/all users to the systemd-coredump group
#**************************************************

while read -r line; do
set -- $(IFS=':' ; echo $line)
usermod --append --groups systemd-coredump $1
done < <(grep /home/ /etc/passwd)

#******************************************************************
# users have to log out/in to pick up the new systemd-coredump group.
# verify user can create a core dump, In a user terminal run
#******************************************************************

firefox &
pkill --signal SIGSEGV --full firefox
ls -Al /var/tmp/
rm /var/tmp/*firefox*.core

#*******************************************************************
# When a user launches a terminal, $HOME/.bashrc is normally executed.
# I can recommend adding a command to check for core files. I also have
# a hourly cron job to warn about core files.
#*******************************************************************

------8<------8<------8<--cut below this line ----8<------8<
#!/bin/bash
#********************* start ck_4_core *******************************

for _d in $HOME /tmp /var/tmp /var/lib/systemd/coredump ; do
_cnt=$(ls $_d/*core* 2> /dev/null | wc -l)
if [ $_cnt -gt 0 ] ; then
echo " "
ls -la $_d/*core*
echo "
# from $(hostname --short) ck_4_core
# There is a $_d/*core file. To remove it, run
rm --force $_d/*core*
"
fi
done

#*************** end ck_4_core ****************************************
------8<------8<------8<--cut above this line ----8<------8<

Do remember to set execute bit on ck_4_core.
chmod +x ck_4_core

#*********** end of enable_coredump.txt ******************************

On 24/06/2021 13:13, Bit Twister wrote:
> On Thu, 24 Jun 2021 12:55:34 +0100, Grimble wrote:
>> On 23/06/2021 17:41, Bit Twister wrote:
>>> On Wed, 23 Jun 2021 14:56:33 +0100, Grimble wrote:
>>>> I noticed several of the 5.10.43
>>>> files in /boot, such as vmlinuz- and sysvers- were 0 B.
>>>
>>> Then those are failures. /boot should look something like this snippet:
>>>
>>> 239K Jun 11 02:33 config-5.10.43-desktop-1.mga8
>>> 4.0K Dec 26 08:34 dracut
>>> 4.0K Jun 14 09:21 grub2
>>> 17M Jun 14 09:12 initrd-5.10.43-desktop-1.mga8.img
>>> 33 Jun 14 09:13 initrd-desktop.img -> initrd-5.10.43-desktop-1.mga8.img
>>> 33 Jun 14 09:13 initrd.img -> initrd-5.10.43-desktop-1.mga8.img
>>> 193K Jun 11 02:33 symvers-5.10.43-desktop-1.mga8.xz
>>> 4.8M Jun 11 02:33 System.map-5.10.43-desktop-1.mga8
>>> 30 Jun 14 09:13 vmlinuz -> vmlinuz-5.10.43-desktop-1.mga8
>>> 7.7M Jun 11 02:33 vmlinuz-5.10.43-desktop-1.mga8
>>> 30 Jun 14 09:13 vmlinuz-desktop -> vmlinuz-5.10.43-desktop-1.mga8
>>>
>> Thanks, Bit. Fortunately 5.10.45 installed without problems.
>
> Well that is a step forward. Since the old kernels did not compile you
> might consider using mcc to remove them.
>
>> However,
>> we're still trying to find the root cause of this instability.
>
> Yes I have been following the thread. Maybe enabling core dump might give
> you a chance to find out why.
<Snipping out all your useful code >
Did all that, launched firefox, got immediate message:
Exiting due to channel error
Crash annotation GraphicsCriticalError Receive IPC close with
reason=AbnormalShutdown (t=1.32075)
So, now I have 550MB of crash data. What to do with it?

--
Grimble
Registered Linux User #450547
Machine 'Bach' running Plasma 5.20.4 on 5.10.45-desktop-2.mga8 kernel.
Mageia release 8 (Official) for x86_64

On Fri, 25 Jun 2021 12:18:50 -0400, Grimble <grimble@nomail.afraid.org> wrote:

> On 24/06/2021 13:13, Bit Twister wrote:
>> On Thu, 24 Jun 2021 12:55:34 +0100, Grimble wrote:
>>> On 23/06/2021 17:41, Bit Twister wrote:
>>>> On Wed, 23 Jun 2021 14:56:33 +0100, Grimble wrote:
>>>>> I noticed several of the 5.10.43
>>>>> files in /boot, such as vmlinuz- and sysvers- were 0 B.
>>>>
>>>> Then those are failures. /boot should look something like this snippet:
>>>>
>>>> 239K Jun 11 02:33 config-5.10.43-desktop-1.mga8
>>>> 4.0K Dec 26 08:34 dracut
>>>> 4.0K Jun 14 09:21 grub2
>>>> 17M Jun 14 09:12 initrd-5.10.43-desktop-1.mga8.img
>>>> 33 Jun 14 09:13 initrd-desktop.img -> initrd-5.10.43-desktop-1.mga8.img
>>>> 33 Jun 14 09:13 initrd.img -> initrd-5.10.43-desktop-1.mga8.img
>>>> 193K Jun 11 02:33 symvers-5.10.43-desktop-1.mga8.xz
>>>> 4.8M Jun 11 02:33 System.map-5.10.43-desktop-1.mga8
>>>> 30 Jun 14 09:13 vmlinuz -> vmlinuz-5.10.43-desktop-1.mga8
>>>> 7.7M Jun 11 02:33 vmlinuz-5.10.43-desktop-1.mga8
>>>> 30 Jun 14 09:13 vmlinuz-desktop -> vmlinuz-5.10.43-desktop-1.mga8
>>>>
>>> Thanks, Bit. Fortunately 5.10.45 installed without problems.
>>
>> Well that is a step forward. Since the old kernels did not compile you
>> might consider using mcc to remove them.
>>
>>> However,
>>> we're still trying to find the root cause of this instability.
>>
>> Yes I have been following the thread. Maybe enabling core dump might give
>> you a chance to find out why.
> <Snipping out all your useful code >
> Did all that, launched firefox, got immediate message:
> Exiting due to channel error
> Crash annotation GraphicsCriticalError Receive IPC close with
> reason=AbnormalShutdown (t=1.32075)
> So, now I have 550MB of crash data. What to do with it?

https://fedoraproject.org/wiki/Debugging_guidelines_for_Mozilla_products#Using_coredumpctl_to_get_backtrace

Are you using wayland?
https://bugzilla.mozilla.org/show_bug.cgi?id=1538435

Regards, Dave Hodgins

--
Change dwhodgins@nomail.afraid.org to davidwhodgins@teksavvy.com for
email replies.

On Fri, 25 Jun 2021 17:18:50 +0100, Grimble wrote:

> Did all that, launched firefox, got immediate message:
> Exiting due to channel error
> Crash annotation GraphicsCriticalError Receive IPC close with
> reason=AbnormalShutdown (t=1.32075)
> So, now I have 550MB of crash data. What to do with it?

Delete it unless you want to just practice on it. The kill command was
just to cause firefox to create a dump and that you can create a core file.

Now all you can do is wait for a crash/core caused by whatever is
causing causing your unstable system problem.

Plug
analyze core file
into a search engine to get hints/commands to work on a core file.

On Fri, 25 Jun 2021 17:18:50 +0100, Grimble wrote:

><Snipping out all your useful code >

I forgot to add information about running ck_4_core in cron.

#*******************************************************************
# If using cron to run ck_4_core you have to monitor root's email.
# If you have postfix installed/running I can recommend that you
# have it send root email to your account. As root run
# cd /etc/postfix
# tail -12 aliases
# # Example of my aliases change
# # diff /var/local/vorig/etc/postfix/aliases_vinstall /etc/postfix/aliases
# 80c80
# < root: postfix
# ---
# > root: bittwister
# # After aliases change run
# postalias aliases
# systemctl restart postfix
# # and send a test message
# mail -s "root email test shot" root < /dev/null
# # Now check that the message is in your Linux mail inbox
# # You may want to do a normal shutdown/startup to see if there are
# any core files generated which you would need to ignore/delete.
# I find lightdm Display Manager core file after each shutdown.
# #*******************************************************************

On 26/06/2021 01:43, Bit Twister wrote:
> On Fri, 25 Jun 2021 17:18:50 +0100, Grimble wrote:
>
>
>> Did all that, launched firefox, got immediate message:
>> Exiting due to channel error
>> Crash annotation GraphicsCriticalError Receive IPC close with
>> reason=AbnormalShutdown (t=1.32075)
>> So, now I have 550MB of crash data. What to do with it?
>
> Delete it unless you want to just practice on it. The kill command was
> just to cause firefox to create a dump and that you can create a core file.
>
Didn't make myelf clear, Bit - I didn't give the Kill command, so it's a
real dump to analyse. Thanks for the tips.

--
Grimble
Registered Linux User #450547
Machine 'Bach' running Plasma 5.20.4 on 5.10.45-desktop-2.mga8 kernel.
Mageia release 8 (Official) for x86_64