Re: postfix fails to start on boot

[Tim via users]

On Thu, 2024-05-16 at 21:40 -0400, Alex wrote:
> Hi,
> I have a fedora38 server with postfix-3.7.9 (although this problem
> has existed for a long time) that fails to start because I believe
> the server has multiple interfaces.
>
> ifconfig shows just the primary ethernet interface, but "ip addr"
> shows the rest. There are five total interfaces in this multi-
> instance postfix config.

Are you using them all?  Are they all up?  Do you want postfix to work
with all of them?  I'm guessing so, but it's not absolutely clear.

If there's a "no" in there, have you tried configuring postfix to only
care about a particular interface or IP?

If "yes," how have you configured postfix to listen?  To specific
interfaces, to a generic "all"?


> After the system boots, I can start postfix successfully with just
> "service postfix start".
> 
> Is it perhaps trying to start before networking is available? This
> would be odd because the problem has persisted for quite a while, and
> I would have thought something like that would have been fixed long
> ago.
> 
> Perhaps related to the difference between ifconfig and ip? Where is
> the egrep coming from? I don't see it in the systemd file:
> 
> # ls -l /etc/systemd/system/multi-user.target.wants/postfix.service
> lrwxrwxrwx 1 root root 39 Jan 24  2023 
> /etc/systemd/system/multi-user.target.wants/postfix.service -> 
> /usr/lib/systemd/system/postfix.service

It could well be down to it trying to start too early, and other people
have dealt with it by themselves, or their hardware comes up quicker
than yours does.

You may want to look into network-online.target, and define what you
consider as the network being up (is one interface adequate or all
interfaces required?).

multi-user.target.wants is rather simple /the OS is running/, and it
can do that without there being a network.

I kinda wish there was a general purpose "non-urgent" parameter for
some services.  They wouldn't try to immediately start as the system
booted, cutting down on the flurry of things trying to fire up.  Let
just the essential services do that.  After that, *then* look at the
remaining ones.

 
-- 
 
uname -rsvp
Linux 3.10.0-1160.118.1.el7.x86_64 #1 SMP Wed Apr 24 16:01:50 UTC 2024 x86_64
 
Boilerplate:  All unexpected mail to my mailbox is automatically deleted.
I will only get to see the messages that are posted to the mailing list.
 
--
___
users mailing list -- users@lists.fedoraproject.org
To unsubscribe send an email to users-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

Re: postfix fails to start on boot

[Chris Adams]

Once upon a time, Alex  said:
> I have a fedora38 server with postfix-3.7.9 (although this problem has
> existed for a long time) that fails to start because I believe the server
> has multiple interfaces.

The issue is that the postfix.service unit has After=network.target
(which is fine for binding to 0.0.0.0 (or :: for IPv6).  When instead
you want to bind to specific IPs (especially beyond just localhost), it
needs to wait for the entire network to be up, which instead needs to be
After=network-online.target.

Run "sudo systemctl edit postfix.service" and insert these lines:

[Unit]
After=network-online.target

That should take care of it.
-- 
Chris Adams 
--
___
users mailing list -- users@lists.fedoraproject.org
To unsubscribe send an email to users-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

Re: postfix fails to start on boot

[Mike Wright]

On 5/16/24 18:40, Alex wrote:

Hi,
I have a fedora38 server with postfix-3.7.9 (although this problem has
existed for a long time) that fails to start because I believe the server
has multiple interfaces.

ifconfig shows just the primary ethernet interface, but "ip addr" shows the
rest. There are five total interfaces in this multi-instance postfix config.

Here are the related boot messages from journalctl -xb0:
postfix[1360]: egrep: warning: egrep is obsolescent; using grep -E
postfix[1371]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.197
postfix[1374]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.198
postfix[1375]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.199
postfix[1376]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.200
postfix[1379]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.201

Each of the IPs correspond to a different postfix instance.

After the system boots, I can start postfix successfully with just "service
postfix start".


Since it has systemd as an option and the postfix service should know to 
wait for networking, why not use that to run it?


systemctl unmask postfix
systemctl enable postfix
systemctl start postfix

After that it should start automatically on boot.



Is it perhaps trying to start before networking is available? This would be
odd because the problem has persisted for quite a while, and I would have
thought something like that would have been fixed long ago.

Perhaps related to the difference between ifconfig and ip? Where is the
egrep coming from? I don't see it in the systemd file:

# ls -l /etc/systemd/system/multi-user.target.wants/postfix.service
lrwxrwxrwx 1 root root 39 Jan 24  2023
/etc/systemd/system/multi-user.target.wants/postfix.service ->
/usr/lib/systemd/system/postfix.service


--
___
users mailing list -- users@lists.fedoraproject.org
To unsubscribe send an email to users-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

postfix fails to start on boot

[Alex]

Hi,
I have a fedora38 server with postfix-3.7.9 (although this problem has
existed for a long time) that fails to start because I believe the server
has multiple interfaces.

ifconfig shows just the primary ethernet interface, but "ip addr" shows the
rest. There are five total interfaces in this multi-instance postfix config.

Here are the related boot messages from journalctl -xb0:
postfix[1360]: egrep: warning: egrep is obsolescent; using grep -E
postfix[1371]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.197
postfix[1374]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.198
postfix[1375]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.199
postfix[1376]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.200
postfix[1379]: fatal: parameter inet_interfaces: no local interface found
for 130.250.NNN.201

Each of the IPs correspond to a different postfix instance.

After the system boots, I can start postfix successfully with just "service
postfix start".

Is it perhaps trying to start before networking is available? This would be
odd because the problem has persisted for quite a while, and I would have
thought something like that would have been fixed long ago.

Perhaps related to the difference between ifconfig and ip? Where is the
egrep coming from? I don't see it in the systemd file:

# ls -l /etc/systemd/system/multi-user.target.wants/postfix.service
lrwxrwxrwx 1 root root 39 Jan 24  2023
/etc/systemd/system/multi-user.target.wants/postfix.service ->
/usr/lib/systemd/system/postfix.service
--
___
users mailing list -- users@lists.fedoraproject.org
To unsubscribe send an email to users-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue

Re: My Fedora 40 experiences

[Stephen Morris]

On 16/5/24 22:13, John Pilkington wrote:

On 16/05/2024 12:33, George N. White III wrote:
On Thu, May 16, 2024 at 6:46 AM Stephen Morris 
mailto:samor...@netspace.net.au>> wrote:


    I used dnf to system-upgrade to F40, which downloaded all the 
relevant

    packages, and then I rebooted through dnf.


Did you do all the steps in 
>? 



    The restart updated all the packages and automatically rebooted 
when it

    was finished.
    When the grub menu was displayed after the reboot it was obvious 
that

    the upgrade did not update the grub config as there was no entry to
    boot
    off the F40 installed kernel.
    I booted to the display manager which I think is still gdm, and the
    first thing I noticed there was the "plasma on xorg" selection 
had been
    removed but the "Gnome on xorg" and "Gnome Classic on xorg" were 
still

    there.
    I selected the "Plasma" entry and booted into Plasma.
    Having loaded Plasma I then went into the system entries menu and 
went

    through all the options again, and it was obvious from this that the
    display options I had configured with F39 had been wiped with F40. I
    configured the display settings again and set the new HDR option 
as I

    have a HDR monitor.
    After the configuration changes I rebuilt the grub menus using
    grub2-mkconfig to get a boot entry for the new kernel, and rebooted.
    With the reboot from the new kernel, and for that matter any of the
    older kernels, I got a message that the nvidia driver was not 
found and

    it was falling back to the nouveau driver, I don't know how as the
    nvidia driver was black listed in the grub menus.


Which nvidia driver was blacklisted (nouveau or nvidia-???)? There 
have been
problems with older Nvidia cards.  Were you using nouveau on F39 or 
nvidia, and
how did you install the nvidia driver (there are multiple sites with 
different installers).
My old iMac uses 470xx from rpmfusion, which was initially in 
no-maintainer status
and required a simple patch to run on current kernels.  Last I check 
the driver was in

testing.

  From the display manager I loaded Plasma as I did after the first
    upgrade boot, and Plasma displayed a black screen and never went any
    further, irrespective of how long I left it for, and the only way I
    could get out of that state was to use the physical reset button 
on the

    computer.
    On reboot, if I selected "Gnome" or "Gnome on Xorg", gnome would 
start
    up quite happily, but logging out and starting Plasma would still 
hang

    the computer.
    So rebooting again to the display manager login screen, I used
    ctrl+alt+F2 to switch to a terminal login process. Logging into the
    terminal the first thing I did was use mokutil to check the uefi 
status
    and it told me that uefi was disabled even though it was enabled 
in the

    bios.
    I checked whether the signing key was enrolled and it was, but the
    system wasn't using the nvidia driver even though it was 
installed, and

    a reinstall of kmod-nvidia and akmod-nvidia did nothing to alleviate
    the
    issue.


Many users have had problems with the akmod-nvida install.  For 470xx
the module failed to compile.  For newer cards, users sometimes end up
with unsigned drivers.  This usually means they rebooted too quickly 
(during

the window after the module was compiled but before it was signed.).

    So I uninstalled the kmod-nvidia module, and did a force 
re-enroll of
    the uefi signed key (potentially with a new key), and then 
rebooted to
    go through the mokutil enrolment required at boot. This did not 
resolve

    the Plasm start issue, so I loaded Gnome.
    Once in Gnome I started Firefox Nightly to do some net searches 
to see
    if I could find a solution, and the one thing that did happen at 
this

    point was the boot did not display the "falling back to nouveau"
    message, but the start of firefox displayed a message that a gpu
    couldn't be found on pci. Looking for a resolution to this I 
found an

    entry about, as part of installed nvidia drivers, to ensure the gpu
    firmware was also installed.
    So I did a dnf install nvidia-gpu-firmware which was then 
installed as

    it hadn't been already.


If you had installed nvidia-??? from a 3rd party repo the Fedora 
firmware

package might not be required, but it is needed for nouveau.

    After installing the firmware I rebooted and started Plasma, which
    successfully started without issues, and when I checked the video
    driver
    it was finally using the nvidia driver.
    This was a lot of work to resolve, and is the worst experience 
I've had

    with any fedora system upgrade.


Blame Nvidia for forcing Fedora into convoluted workarounds instead of
helping make older hardware work on current kernels.

Re: My Fedora 40 experiences

[Stephen Morris]

On 16/5/24 21:33, George N. White III wrote:
On Thu, May 16, 2024 at 6:46 AM Stephen Morris 
 wrote:


I used dnf to system-upgrade to F40, which downloaded all the
relevant
packages, and then I rebooted through dnf.


Did you do all the steps in 
?

I did, except step 4 as I wasn't asked to verify the imported key.


The restart updated all the packages and automatically rebooted
when it
was finished.
When the grub menu was displayed after the reboot it was obvious that
the upgrade did not update the grub config as there was no entry
to boot
off the F40 installed kernel.
I booted to the display manager which I think is still gdm, and the
first thing I noticed there was the "plasma on xorg" selection had
been
removed but the "Gnome on xorg" and "Gnome Classic on xorg" were
still
there.
I selected the "Plasma" entry and booted into Plasma.
Having loaded Plasma I then went into the system entries menu and
went
through all the options again, and it was obvious from this that the
display options I had configured with F39 had been wiped with F40. I
configured the display settings again and set the new HDR option as I
have a HDR monitor.
After the configuration changes I rebuilt the grub menus using
grub2-mkconfig to get a boot entry for the new kernel, and rebooted.
With the reboot from the new kernel, and for that matter any of the
older kernels, I got a message that the nvidia driver was not
found and
it was falling back to the nouveau driver, I don't know how as the
nvidia driver was black listed in the grub menus.


Which nvidia driver was blacklisted (nouveau or nvidia-???)?  There 
have been
problems with older Nvidia cards.  Were you using nouveau on F39 or 
nvidia, and
how did you install the nvidia driver (there are multiple sites with 
different installers).
My old iMac uses 470xx from rpmfusion, which was initially in 
no-maintainer status
and required a simple patch to run on current kernels. Last I check 
the driver was in

testing.
Sorry, I meant the nouveau driver was blacklisted. I have always used 
the nvidia driver from rpmfusion as I have needed the hardware 
acceleration the nvidia driver provides. I was using the nvidia driver 
in F39 both via kmod-nvidia and akmod-nvidia for backup.


 From the display manager I loaded Plasma as I did after the first
upgrade boot, and Plasma displayed a black screen and never went any
further, irrespective of how long I left it for, and the only way I
could get out of that state was to use the physical reset button
on the
computer.
On reboot, if I selected "Gnome" or "Gnome on Xorg", gnome would
start
up quite happily, but logging out and starting Plasma would still
hang
the computer.
So rebooting again to the display manager login screen, I used
ctrl+alt+F2 to switch to a terminal login process. Logging into the
terminal the first thing I did was use mokutil to check the uefi
status
and it told me that uefi was disabled even though it was enabled
in the
bios.
I checked whether the signing key was enrolled and it was, but the
system wasn't using the nvidia driver even though it was
installed, and
a reinstall of kmod-nvidia and akmod-nvidia did nothing to
alleviate the
issue.


Many users have had problems with the akmod-nvida install.  For 470xx
the module failed to compile.  For newer cards, users sometimes end up
with unsigned drivers.  This usually means they rebooted too quickly 
(during

the window after the module was compiled but before it was signed.).
I have had the reboot happen too too quickly before but in this case I 
had no control over the reboot process, it happened automatically when 
the installs were completed. I'm using the 550 driver as that is one the 
system selected for my RTX 3080 graphics card.


So I uninstalled the kmod-nvidia module, and did a force re-enroll of
the uefi signed key (potentially with a new key), and then
rebooted to
go through the mokutil enrolment required at boot. This did not
resolve
the Plasm start issue, so I loaded Gnome.
Once in Gnome I started Firefox Nightly to do some net searches to
see
if I could find a solution, and the one thing that did happen at this
point was the boot did not display the "falling back to nouveau"
message, but the start of firefox displayed a message that a gpu
couldn't be found on pci. Looking for a resolution to this I found an
entry about, as part of installed nvidia drivers, to ensure the gpu
firmware was also installed.
So I did a dnf install nvidia-gpu-firmware which was then
installed as
it hadn't been already.


If you had installed nvidia-??? from a 3rd party repo the Fedora firmware
package might not be required, 

Re: My Fedora 40 experiences

[John Pilkington]

On 16/05/2024 12:33, George N. White III wrote:
On Thu, May 16, 2024 at 6:46 AM Stephen Morris > wrote:


I used dnf to system-upgrade to F40, which downloaded all the relevant
packages, and then I rebooted through dnf.


Did you do all the steps in 
>?


The restart updated all the packages and automatically rebooted when it
was finished.
When the grub menu was displayed after the reboot it was obvious that
the upgrade did not update the grub config as there was no entry to
boot
off the F40 installed kernel.
I booted to the display manager which I think is still gdm, and the
first thing I noticed there was the "plasma on xorg" selection had been
removed but the "Gnome on xorg" and "Gnome Classic on xorg" were still
there.
I selected the "Plasma" entry and booted into Plasma.
Having loaded Plasma I then went into the system entries menu and went
through all the options again, and it was obvious from this that the
display options I had configured with F39 had been wiped with F40. I
configured the display settings again and set the new HDR option as I
have a HDR monitor.
After the configuration changes I rebuilt the grub menus using
grub2-mkconfig to get a boot entry for the new kernel, and rebooted.
With the reboot from the new kernel, and for that matter any of the
older kernels, I got a message that the nvidia driver was not found and
it was falling back to the nouveau driver, I don't know how as the
nvidia driver was black listed in the grub menus.


Which nvidia driver was blacklisted (nouveau or nvidia-???)?  There have 
been
problems with older Nvidia cards.  Were you using nouveau on F39 or 
nvidia, and
how did you install the nvidia driver (there are multiple sites with 
different installers).
My old iMac uses 470xx from rpmfusion, which was initially in 
no-maintainer status
and required a simple patch to run on current kernels.  Last I check the 
driver was in

testing.

  From the display manager I loaded Plasma as I did after the first
upgrade boot, and Plasma displayed a black screen and never went any
further, irrespective of how long I left it for, and the only way I
could get out of that state was to use the physical reset button on the
computer.
On reboot, if I selected "Gnome" or "Gnome on Xorg", gnome would start
up quite happily, but logging out and starting Plasma would still hang
the computer.
So rebooting again to the display manager login screen, I used
ctrl+alt+F2 to switch to a terminal login process. Logging into the
terminal the first thing I did was use mokutil to check the uefi status
and it told me that uefi was disabled even though it was enabled in the
bios.
I checked whether the signing key was enrolled and it was, but the
system wasn't using the nvidia driver even though it was installed, and
a reinstall of kmod-nvidia and akmod-nvidia did nothing to alleviate
the
issue.


Many users have had problems with the akmod-nvida install.  For 470xx
the module failed to compile.  For newer cards, users sometimes end up
with unsigned drivers.  This usually means they rebooted too quickly 
(during

the window after the module was compiled but before it was signed.).

So I uninstalled the kmod-nvidia module, and did a force re-enroll of
the uefi signed key (potentially with a new key), and then rebooted to
go through the mokutil enrolment required at boot. This did not resolve
the Plasm start issue, so I loaded Gnome.
Once in Gnome I started Firefox Nightly to do some net searches to see
if I could find a solution, and the one thing that did happen at this
point was the boot did not display the "falling back to nouveau"
message, but the start of firefox displayed a message that a gpu
couldn't be found on pci. Looking for a resolution to this I found an
entry about, as part of installed nvidia drivers, to ensure the gpu
firmware was also installed.
So I did a dnf install nvidia-gpu-firmware which was then installed as
it hadn't been already.


If you had installed nvidia-??? from a 3rd party repo the Fedora firmware
package might not be required, but it is needed for nouveau.

After installing the firmware I rebooted and started Plasma, which
successfully started without issues, and when I checked the video
driver
it was finally using the nvidia driver.
This was a lot of work to resolve, and is the worst experience I've had
with any fedora system upgrade.


Blame Nvidia for forcing Fedora into convoluted workarounds instead of
helping make older hardware work on current kernels.

Rpmfusion is having difficulties finding volunteers to package software.
Fewer 

Re: My Fedora 40 experiences

[George N. White III]

On Thu, May 16, 2024 at 6:46 AM Stephen Morris 
wrote:

> I used dnf to system-upgrade to F40, which downloaded all the relevant
> packages, and then I rebooted through dnf.
>

Did you do all the steps in <
https://docs.fedoraproject.org/en-US/quick-docs/upgrading-fedora-offline/>?


> The restart updated all the packages and automatically rebooted when it
> was finished.
> When the grub menu was displayed after the reboot it was obvious that
> the upgrade did not update the grub config as there was no entry to boot
> off the F40 installed kernel.
> I booted to the display manager which I think is still gdm, and the
> first thing I noticed there was the "plasma on xorg" selection had been
> removed but the "Gnome on xorg" and "Gnome Classic on xorg" were still
> there.
> I selected the "Plasma" entry and booted into Plasma.
> Having loaded Plasma I then went into the system entries menu and went
> through all the options again, and it was obvious from this that the
> display options I had configured with F39 had been wiped with F40. I
> configured the display settings again and set the new HDR option as I
> have a HDR monitor.
> After the configuration changes I rebuilt the grub menus using
> grub2-mkconfig to get a boot entry for the new kernel, and rebooted.
> With the reboot from the new kernel, and for that matter any of the
> older kernels, I got a message that the nvidia driver was not found and
> it was falling back to the nouveau driver, I don't know how as the
> nvidia driver was black listed in the grub menus.
>

Which nvidia driver was blacklisted (nouveau or nvidia-???)?  There have
been
problems with older Nvidia cards.  Were you using nouveau on F39 or nvidia,
and
how did you install the nvidia driver (there are multiple sites with
different installers).
My old iMac uses 470xx from rpmfusion, which was initially in no-maintainer
status
and required a simple patch to run on current kernels.  Last I check the
driver was in
testing.


>  From the display manager I loaded Plasma as I did after the first
> upgrade boot, and Plasma displayed a black screen and never went any
> further, irrespective of how long I left it for, and the only way I
> could get out of that state was to use the physical reset button on the
> computer.
> On reboot, if I selected "Gnome" or "Gnome on Xorg", gnome would start
> up quite happily, but logging out and starting Plasma would still hang
> the computer.
> So rebooting again to the display manager login screen, I used
> ctrl+alt+F2 to switch to a terminal login process. Logging into the
> terminal the first thing I did was use mokutil to check the uefi status
> and it told me that uefi was disabled even though it was enabled in the
> bios.
> I checked whether the signing key was enrolled and it was, but the
> system wasn't using the nvidia driver even though it was installed, and
> a reinstall of kmod-nvidia and akmod-nvidia did nothing to alleviate the
> issue.
>

Many users have had problems with the akmod-nvida install.  For 470xx
the module failed to compile.  For newer cards, users sometimes end up
with unsigned drivers.  This usually means they rebooted too quickly
(during
the window after the module was compiled but before it was signed.).


> So I uninstalled the kmod-nvidia module, and did a force re-enroll of
> the uefi signed key (potentially with a new key), and then rebooted to
> go through the mokutil enrolment required at boot. This did not resolve
> the Plasm start issue, so I loaded Gnome.
> Once in Gnome I started Firefox Nightly to do some net searches to see
> if I could find a solution, and the one thing that did happen at this
> point was the boot did not display the "falling back to nouveau"
> message, but the start of firefox displayed a message that a gpu
> couldn't be found on pci. Looking for a resolution to this I found an
> entry about, as part of installed nvidia drivers, to ensure the gpu
> firmware was also installed.
> So I did a dnf install nvidia-gpu-firmware which was then installed as
> it hadn't been already.
>

If you had installed nvidia-??? from a 3rd party repo the Fedora firmware
package might not be required, but it is needed for nouveau.


> After installing the firmware I rebooted and started Plasma, which
> successfully started without issues, and when I checked the video driver
> it was finally using the nvidia driver.
> This was a lot of work to resolve, and is the worst experience I've had
> with any fedora system upgrade.
>

Blame Nvidia for forcing Fedora into convoluted workarounds instead of
helping make older hardware work on current kernels.

Rpmfusion is having difficulties finding volunteers to package software.
Fewer volunteers have access to older hardware and those that do are
getting old themselves and may be dealing with the problems facing older
people in our society, so rpmfusion falls down in their priority list.

-- 
George N. White III
--
___
users 

My Fedora 40 experiences

[Stephen Morris]

I used dnf to system-upgrade to F40, which downloaded all the relevant 
packages, and then I rebooted through dnf.
The restart updated all the packages and automatically rebooted when it 
was finished.
When the grub menu was displayed after the reboot it was obvious that 
the upgrade did not update the grub config as there was no entry to boot 
off the F40 installed kernel.
I booted to the display manager which I think is still gdm, and the 
first thing I noticed there was the "plasma on xorg" selection had been 
removed but the "Gnome on xorg" and "Gnome Classic on xorg" were still 
there.

I selected the "Plasma" entry and booted into Plasma.
Having loaded Plasma I then went into the system entries menu and went 
through all the options again, and it was obvious from this that the 
display options I had configured with F39 had been wiped with F40. I 
configured the display settings again and set the new HDR option as I 
have a HDR monitor.
After the configuration changes I rebuilt the grub menus using 
grub2-mkconfig to get a boot entry for the new kernel, and rebooted.
With the reboot from the new kernel, and for that matter any of the 
older kernels, I got a message that the nvidia driver was not found and 
it was falling back to the nouveau driver, I don't know how as the 
nvidia driver was black listed in the grub menus.
From the display manager I loaded Plasma as I did after the first 
upgrade boot, and Plasma displayed a black screen and never went any 
further, irrespective of how long I left it for, and the only way I 
could get out of that state was to use the physical reset button on the 
computer.
On reboot, if I selected "Gnome" or "Gnome on Xorg", gnome would start 
up quite happily, but logging out and starting Plasma would still hang 
the computer.
So rebooting again to the display manager login screen, I used 
ctrl+alt+F2 to switch to a terminal login process. Logging into the 
terminal the first thing I did was use mokutil to check the uefi status 
and it told me that uefi was disabled even though it was enabled in the 
bios.
I checked whether the signing key was enrolled and it was, but the 
system wasn't using the nvidia driver even though it was installed, and 
a reinstall of kmod-nvidia and akmod-nvidia did nothing to alleviate the 
issue.
So I uninstalled the kmod-nvidia module, and did a force re-enroll of 
the uefi signed key (potentially with a new key), and then rebooted to 
go through the mokutil enrolment required at boot. This did not resolve 
the Plasm start issue, so I loaded Gnome.
Once in Gnome I started Firefox Nightly to do some net searches to see 
if I could find a solution, and the one thing that did happen at this 
point was the boot did not display the "falling back to nouveau" 
message, but the start of firefox displayed a message that a gpu 
couldn't be found on pci. Looking for a resolution to this I found an 
entry about, as part of installed nvidia drivers, to ensure the gpu 
firmware was also installed.
So I did a dnf install nvidia-gpu-firmware which was then installed as 
it hadn't been already.
After installing the firmware I rebooted and started Plasma, which 
successfully started without issues, and when I checked the video driver 
it was finally using the nvidia driver.
This was a lot of work to resolve, and is the worst experience I've had 
with any fedora system upgrade.


regards,
Steve


OpenPGP_0x594338B1DE179AB2.asc
Description: OpenPGP public key


OpenPGP_signature.asc
Description: OpenPGP digital signature
--
___
users mailing list -- users@lists.fedoraproject.org
To unsubscribe send an email to users-le...@lists.fedoraproject.org
Fedora Code of Conduct: 
https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: 
https://lists.fedoraproject.org/archives/list/users@lists.fedoraproject.org
Do not reply to spam, report it: 
https://pagure.io/fedora-infrastructure/new_issue