Pfsese 2.6ce installation fails to asus z690 motheboard
-
Board - Asus TUF GAMING Z690-PLUS WIFI D4
cpu - intel 12700K
ram - 32GBThis is my test system and completely bare bones. Using onboard video.
Attempting to install from a flash drive, the installer boots up to the screen below then locks.
Then just locks.
Earlier I see something relating to a serial port. The onboard serial port however is disabled.
Enabling it and changing the address/irq assignment seems to allow the installer to continue booting and actually get to the point where either the live version or installer option becomes available.
There is no physical serial port installed to the motherboard header. It's unclear why the OS is even detecting anything related to a serial port when disabled in the bios.
I tried both the vga and dvd medias with same result. Sure I can leave the serial port enabled in the bios, but would like to understand why this is happening.
Thoughts?
-
When the installer boots, check what it shows for the console. Make sure it says video before continuing to boot. In some cases it may be trying both consoles or preferring serial there. You can hit the loader menu option to cycle through the choices.
Hardware isn't usually that picky but it's third party gear on FreeBSD so anything is possible.
-
@jimp Thank you for your reply.
I forgot to mention, I did try adjusting the console output to vga only, with same result. It almost seems like setting serial to disabled in bios doesn't really disable it.
One other tidbit I forgot to mention, with the serial port enabled, on address #3F8, irq 4 in the bios, installer boots but gets a blank screen. This suggests it is trying to output to serial. I will retest various combo's and report back.
-
You're better off leaving the serial port enabled, there isn't going to be a benefit to shutting it down.
It's also possible there is a quirk with that new hardware and FreeBSD 12.x, so you may want to try installing a 2.7 snapshot to see if that succeeds.
Another thing you might try is that if it's booting EFI, try having it boot legacy mode to see if that works, assuming the BIOS has an option for that.
-
https://www.asus.com/us/support/FAQ/1045467/
Appears after introduction of the intel 500 series chipset, csm functionality is only possible with dgpu installed. Integrated video is not supported with csm.
It's unclear if this is an asus quirk or affects other board manufactures as well. One would think if this is a byproduct of the chipset, it should affect all boards using that series.
2.7 snapshot has the same issue re serial port.
<mind blown>
-
@jimp Added a video card so I could enable CSM. Confirmed console set to video (not serial). Same result.
This issue is not the end of the world, but quite bizarre indeed.
-
I gave up on disabling the serial port and just let it be.
Installed 2.6.0 successfully then updated to 22.05+
Things got hairy after this. Trying to update to 23.01 resulted in massive file corruption and unbootable system. I'm using efi/UFS file system, not zfs.
Doing more research I stumbled onto this
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=261169
The tl;dr apparently there's incompatibility with the P & E core arrangement starting with freebsd 13. The advised fix was to disable the E cores (which is possible in the bios). Did that, reinstalled 2.6.0, updated to 22.05 then to 23.05. It booted 23.05 successfully with no corruption.
Not sure if any of the netgate products are using any alderlake (12th gen) or newer processors which have the p & e cores.
-
I see from that issue some fixes have made it into FreeBSD but after our last src merge for 23.01. We should have those in the next release, 23.05, so you should (in theory) be able to re-enable them eventually. Hard to know without trying, though.
-
@gpz1100 said in Pfsese 2.6ce installation fails to asus z690 motheboard:
Not sure if any of the netgate products are using any alderlake (12th gen) or newer processors which have the p & e cores.
Nope, as that are mostly consumer or specific (app)server hardware. Netgate is primarily running Atom-C and Xeon-D hardware only currently (network SOC) - besides the ARM boxes of course. Don't really see the benefit of the mixed architecture on server/network equipment as you normally don't have that much things you'd delegate to an E-core. I could only see that with some sort of fixed bind that would e.g. put specific jobs to Es but otherwise block them from getting any tasks. In networks you're normally very dependent on low latency and fast throughput so a system wouldn't really use a slow core intentionally as its whole goal is to put things through ASAP. For pfSense the only thing that would jump to my mind would be the WebUI, Log handling or backup tasks that could make use of these. All other things you don't want to deliberately slow down your processing speed.
I can see that working in a hypervisor setup though, with the E-cores running the HV/management stack and the P-cores dedicated to the routing VM. That would make sense, but that's exactly the use case as it's specific to an (app)server or in that case hypervisor that can make good use of those. :)
Nevertheless interesting findings that there are quirks about running that mix on baremetal! Thanks for that!
Cheers
\jens -
FYI- The CE 2.7 snapshot builds from today should be using a newer merge from main that should have the fixes for this. It's worth trying again if you have the time to do so.
-
@jimp Tested the 3/1/2023 release of CE 2.7 with E core enabled. No corruption and it completed the boot cycle successfully.
