CentOS9, Red hat 9 really slow boot and not display on build 20486664

[fabienmagagnosc]

I just upgraded VMware fusion to the latest version 20486664, and the boot of all the Fedora/CentOS/Red Hat on kernal 5.14.x are simply so long, we speak about nearly 5 minutes on a M1 Max, with 4 cores allocated and 16GB of RAM !

Else, the previous version was really fast ... but we got Ubuntu now 😉

[Technogeezer]

That lengthy boot time has not been my experience on this version of the Tech Preview with any of the Red Hat family distros.  I have built RHEL 9, CentOS 9 Stream, and Fedora 35/36/37 VMs. 

What version of the Tech Preview did you upgrade from.

Can you be a bit more specific on which versions of those distros, and the exact kernel versions involved?

Also, are your VMs configured for NAT or bridged networking? 

[Technogeezer]

I just ran a couple of experiments on my CentOS 9 VM, running the latest Tech Preview 22H2 build.

The kernel that I had installed in my CentOS 9 Stream VM when I booted it was 5.14.0-134.el9. (I haven't started that VM for a while). That kernel booted quite snappily and displays messages to the console during the boot.

Since I hadn't updated the VM in a while, so I decided to run an update. The update installed a new kernel 5.14.0-171. This kernel appears slower to boot, and does not display any messages as it boots. The screen is blank screen until the graphical login appears. A distinct change in prior behavior.

Switching back to the 5.14.0-134 kernel in the GRUB boot menu, and the kernel returns to booting "snappily".

It does appear that CentOS has changed something in their latest kernel that makes it behave differently. I'm suspect we're seeing the same things. Blacklisting the VMware graphics driver (vmwgfx) does not change the behavior.

I do wish that CentOS and Red Hat would deliver a newer kernel than 5.14. IMO 5.14 is just adequate on arm64 architectures. I'm finding that 5.19 and 6.0 kernels behave much better (and fix kernel bugs that VMware discovered). 

[fabienmagagnosc]

I finally start to have some output, and we can "narrow down" to something happening before the Random Number Generation initialization it seems

it seems that for almost a minutes, "nothing" is happening.

[Technogeezer]

Things seem to have gotten a lot better for me.

I'm now running Fusion 13 with RHEL 9.1 and the latest CentOS 9 Stream distros. I think the kernels have gotten better on arm64.

CentOS 9 stream is now showing a 5.14.0-210.el9 kernel and seems to boot more quickly. No need to blacklist the vmwgfx driver either.

Same with RHEL 9.1 - a 5.14.0-162.5.1.el9_1 kernel with no vmwgfx blacklisting.