CentOS 8.2 Can't boot from latest kernel from disk, or from live usb

[shavera]

I have a Dell Precision P5530 laptop that had CentOS 8.1 installed on it. I'm new to CentOS and I didn't realize that the software manager 'update software' would update the minor version of CentOS or the Kernel, and it didn't seem obvious to me that's what the update was doing. When I tried to boot back up after this, booting from grub, the 8.2 4.18.0-193.6.3 kernel would go to a black screen immediately following. Selecting one of the other options (the 4.18.0-147.8.1 kernel), I was able to successfully boot.

In the successful boot, I enabled persistent boot logging, turned off grub 'quiet' etc. to debug the boot.

I rebooted, attempting the new grub entry again, and still black screen. Booted into the old kernel, there was no log at all from the attempt to boot into the new kernel. Additionally, while there were all of the boot messages booting into the old kernel (and the tux penguins at the top, etc.), there were no messages at all when trying to boot into the new kernel. So it feels like it has to be something very early in the boot process.

I used dracut to try to rebuild the initramfs for the 193 kernel, but still no luck.

I considered that the 8.2 upgrade has some secure linux enhancements, so I tried enabling secure boot in UEFI. When I do this, I get one line "EFI stub: UEFI Secure Boot is enabled." I've poked around at some settings in this area, but I don't know enough to figure that part out.

So I decided I could just reinstall using a liveUSB. First I used the 'boot' iso, then later the 'dvd1' iso. I checked the hash on both files and they matched. I used rufus (from my windows machine) to burn the isos to a usb. However, when I boot from the usb, I actually get the same situation as I do from booting from disk. I can select which mode I want to boot into "Install" "Test Media and Install" etc. I've tried all the modes, and same situation. Black screen if Secure Boot is off, 'EFI Stub' if Secure Boot is on, but nothing else.

The hard disk is encrypted (LUKS) and it isn't asking for the encryption password, so I know it's not getting far enough in the boot where this is like a graphics driver or desktop manager failure where I can get into a terminal via ctrl alt F3, e.g.

The fact that it can't boot from the live USB makes me feel like there's some weird conflict with the kernel in 8.2 and something in the laptop hardware. But I don't know what it could be. In the old kernel successful boot, the wireless connection was flicking on and off constantly, so I tried disabling radios, but still no luck.

Is there anything else I can do to gain some kind of insight into what's failing? There's no boot logs, no grub messages, so I don't even know what to try next.

[TrevorH]

Rufus breaks our iso images with some options selected so it's not on our safe-to-use list. It _can_ work but only if you incant the correct spell so it's generally easier to use something that works all the time. See my sig below for a link to the wiki page about it.

Your first step with problems like this should be to use the grub menu to edit the kernel command line and remove the two options that hide all useful information from you. These are 'rhgb' and 'quiet' so reboot, get the grub menu, edit the latest entry, remove rhgb and quiet, continue the boot. Now you should see reams of kernel messages and it might actually tell you what the problem is.

This sounds like a video problem to me. Do you have multiple video cards in this laptop? Did you ever install proprietary video drivers from nvidia/AMD?

[shavera]

Okay, I can try again with a recommended usb creator (or just dd).

So I tried the grub menu edit, Disabled rhgb and quiet, and absolutely nothing is printed to the screen at all. The messages do print on the older kernel boot, but nothing for the 193 kernel. I even went into grub command line and tried entering all the lines manually, to see if somehow there might be additional logs with that approach and still nothing.

Edit/Update: Used `dd`from linux to create the live usb image. Same results. After selecting an item in the initial menu when booting from the USB, I only see a black screen, no other messages, Ctrl Alt F<n> doesn't bring up any terminals.

[dgnzcn]

which graphics card brand do you have AMD / Nvidia?

[mathog]

For whatever it is worth, I recently rebuilt a MythTV box and a Ryzen 3 3200G CPU where I was unable to get the amd_gpu module to load/work with any 4.x kernel. It was necessary to go all the way to an alternate 5.3 kernel (this was Ubuntu 18.04). That said, with the older kernel the "radeon" driver always loaded, and there was never any graphics problem while the kernel booted. Only X11 was affected.

Perhaps this kernel is negotiating a resolution with the display that the latter cannot actually use? Try looking up the supported modes and set parameters for the kernel in grub to use one of them. Or better yet, if your display shows it, boot the older kernel, note the resolution, and use that.

Post the kernel parameters which are being used for the older and newer kernels, perhaps the difference will tell us something.

[shavera]

Intel processor (i7-8850H) and NVidia GPU (Quadro P2000)

I did try (prior to this) to install the NVidia graphics driver (following instructions about disabling nouveau, etc.), but the system only sees the Intel integrated graphics. Since I'm not doing anything too graphically intense, it hasn't been a priority to fix this.

The idea about resolution issues could be an interesting one to look into. The wifi card is being really flaky on the old kernel after the update so I've been getting ethernet via a dell thunderbolt dock. I've tried booting without it plugged in of course, but maybe that could be something. I do know that the integrated screen is 1920x1080.

[dgnzcn]

Intel processor (i7-8850H) and NVidia GPU (Quadro P2000)

I did try (prior to this) to install the NVidia graphics driver (following instructions about disabling nouveau, etc.), but the system only sees the Intel integrated graphics. Since I'm not doing anything too graphically intense, it hasn't been a priority to fix this.

The idea about resolution issues could be an interesting one to look into. The wifi card is being really flaky on the old kernel after the update so I've been getting ethernet via a dell thunderbolt dock. I've tried booting without it plugged in of course, but maybe that could be something. I do know that the integrated screen is 1920x1080.

Please start a recovery option on centos 8 boot menu then run this command:
dracut --regenerate-all --force --install=/lib/firmware/nvidia/*

than reboot

i hope this helps.

[shavera]

From the old kernel booted into rescue mode, I ran

Code: Select all

dracut --regenerate-all --force --install=/lib/firmware/nvidia/*

but the system is still in the same state. Given that I get the same situation when booting from USB, I feel like the problem has to be somewhere earlier in the boot process where it doesn't matter whether it's booting from USB or disk, it can't proceed with the boot.

[shavera]

Through some experimentation, we have been able to boot into the 8.2 live USB and into installed CentOS 8 systems (when upgraded from 8.1 to 8.2) by enabling Legacy BIOS support and not UEFI. When UEFI is reenabled, cannot boot into 8.2 on either system.

Additionally, with Legacy support enabled, and attempting to install directly from an 8.2 USB, the installation fails to install the bootloader (and it doesn't appear clear to me why it fails to install the bootloader)

We've observed this on multiple Dell laptops at this point, so it feels like there must be some conflict between 8.2 UEFI booting and the Dell hardware we're booting on.