I’m trying to install CentOS 7.7.1908 on an Asus X570 TUF Gaming Pro WiFi motherboard with an AMD Ryzen 9 5950X and I am getting a kernel panic issue on it when I plug in the USB installation drive.
I downloaded the ISO from vault.centos.org and wrote it to the USB stick (Patriot 16 GB) using Rufus 3.8.
This is the kind of screen that I get when I try to click on “Install CentOS” when trying to boot from said USB stick:
(The picture isn’t mine but I get something that is very similar to it. That picture comes from someone else from the CentOS forums. I can’t take a picture of my system showing the same kind of kernel panic error because I am currently using Ubuntu 21.10 and it’s calculating parity for me on a really large file, but I would like to switch back over to CentOS if at all possible.)
Has anybody encountered this or something like this before?
Hardware specs:
Processor: AMD Ryzen 9 5950X
Motherboard: Asus X570 TUF Gaming Pro WiFi
RAM: 2x 32 GB Crucial DDR4-3200 CT32G4DFD832A Non-ECC, Unbuffered
Hard drive: HGST 1 TB SATA 6 Gbps 7200 rpm
Video card: EVGA GTX 980
Power supply: Corsair CX750M
OS (attempted): CentOS 7.7.1908
Additional information:
I am using CentOS 7.7.1908 due to HPC/CAE application compatibility issues/reasons. This is my hard limiting factor in terms of why I can’t just switch to any of the newer distros or even newer versions CentOS.
This is likely a failure due to running a 2+ year old kernel on hardware that was released after the kernel. Not only was Ryzen 5000 not released, but Ryzen 3000 was not even out prior to 7.7.1908’s release.
Without seeing the full kernel dump, I can’t say for sure.
I hesitate to blindly start throwing out change suggestions, but if you have other hardware at your disposal that you can boot the liveusb on, which is older than the ISO date (august 2019), it would be a good idea to try booting it on there, to see if you get the same panic. If you get the same panic, re-download the ISO, verify it’s checksum, then reflash it. Sometimes a bit flips here or there in a download. It’s rare, but it is known to happen.
Another option would be to verify your hardware is working properly. I’m not sure if this is a fresh build. You could download the latest Ubuntu iso and try booting that, not to install but again, to see if the kernel panics persist.
My suspicion is that OP’s version requirement is pinned to core lib versions and or kernel version. which makes Alma and Rocky a question mark at best.
So…I actually tried to see if I can “jumpstart” the system by installing the OS on the hard drive using my Intel Core i7-3930K (on an Asus X79 Sabertooth Motherboard) just to get it going (it worked for my 12900K), but alas, to no avail as it also resulted in the kernel panic once I moved the hard drive over back to the Ryzen 5950X system.
So it’s strange that this method worked for the 12900K which, per your comment about the OS/kernel being relatively significantly older than the processors – worked for one but not the other.
(Unfortunately, I don’t have an older AMD system that I can do the same with, so it is entirely possible that this might be why installing on a 3930K - I can’t move said hard drive over to the 5950X.)
That being said, if there is a way for me to install the OS on the 3930K, and then like patch or update/upgrade the kernel and then move the hard drive again, I would be willing to try that.
I’ll try and see if I can get that for you once the 5950X is finished doing what it is currently doing.
A kernel is a kernel. Linux isn’t like Windows in that it throws a fit when moved to another system. Its very likely that the updated kernel simply doesn’t exist in the repositories to support this hardware.
I wonder if bootstrapping it and building a custom newer kernel would work?
Yeah. I didn’t know how much of the kernel is tied to the CPU vs. not. (Features, instructions, optimisations, etc.)
I mean, I can try looking into doing something like that (bootstrapping it to my 3930K, updating the kernel, and then moving the hard drive back over).
If it works, that would be great!
But if it doesn’t, then I’m no worse off than where I am today (with trying to run Ubuntu 21.10).
I have zero experience updating a kernel in CentOS 7.7.1908 (that doesn’t come from the YUM repo) but I reckon that I should be able to google that.
Thanks.
edit
That actually ended up being quite a bit easier than I thought that it was going to be.
I was able to successfullly update the kernel whilst the hard drive was plugged into my 3930K to the 5.15 kernel courtesy of elrepo and then I was able to move the hard drive over to the Ryzen 5950X system and it booted up without any issues.
Yeah…this has been a complete and utter nightmare.
With an AMD Ryzen 9 5950X, if I DON’T update the kernel using elrepo, then the system will have a kernel panic, pretty much “by default” with the default kernel that ships with both CentOS 7.7.1908 (kernel 3.10.0-1062) and CentOS 7.9.2009 (kernel 3.10.0-1160).
So, this means that I pretty much “HAVE TO” update the kernel.
If I update to the elrepo’s mainline kernel (as of this writing 5.15.11), then the 5950X won’t kernel panic, but the “inbox” infiniband drivers will cause a kernel panic.
And if I try to use Mellanox’s MLNX_OFED_LINUX-5.5-1.0.3.2-rhel7.9-x86_64 drivers, and my Mellanox ConnectX-4 MCX456A-ECAT NIC already has the latest firmware flashed to it, then openibd will fail to restart after the installation of said drivers (so that I will stop using said “inbox” drivers which is causing the kernel panic) and/or that the Mellanox card won’t start up properly because I can’t build the kernel modules to support the updated elrepo kernels.
I think that the kernel update is required in order to get CentOS up and running with a Ryzen 5950X.
The problem that I run into with this requirement is that then the Mellanox Infiniband kernel modules aren’t built for that kernel, and therefore; if the source isn’t available, then the Mellanox driver won’t be able to update/build the new kernel modules for whatever kernel it gets updated to, so I’m still without Infiniband on said 5950X in CentOS.
But I’ll try it.
The other thing that I am going to try as well is to see if I can update JUST the CentOS “inbox” Infiniband drivers in the hopes that maybe updating that will be sufficient to alleviate this kernel panic issue with the default/original Infiniband inbox drivers that ships with the OS.
For those that might be interested, here’s where this stands right now:
I did get the mainline kernel 5.14.15 working on my system reliably.
Mainline kernel 5.15.11 will and does produce a repeatable kernel panic with Infiniband and it is NOT restricted to the AMD Zen3/X570 platform.
The same mainline kernel on the 12th Intel Alder Lake/Z690 platform, also with Infiniband will also produce the same kernel panic.
As such, I have filed the bug report to kernel.org.
That being said, as I mentioned, I did get kernel mainline 5.14.15 up and running and it has been stable.
Kernel module k10temp works and so does everything else EXCEPT the onboard Intel I225-V 2.5 GbE NIC. I’m using an Intel Gigabit CT Desktop GbE NIC AIC right now because that works more reliably (based on the Intel 82574L chipset) than said Intel I225-V (at least on Linux).
Course, it’s probably quite rare, if ever, that people would be using Infiniband on a Ryzen system in the first place anyways, but if you do, this is what you can look out for.
