Boot failure after upgrade to 11.4 BTX halted

J

jwdevel

I am doing a source upgrade from 11.3 to 11.4. Now that I have installed the new userland, reboot is failing.

Roughly, these are the steps I have done so far (as I have done on previous updates; nothing new here):
  1. get latest source from SVN
  2. make buildworld
  3. make kernel
  4. reboot - this was succesful. I saw the new kernel active, with the old userland
  5. in single-user mode:
    1. adjkerntz -i
    2. mergemaster -p
    3. make installworld
    4. reboot
And at this point, the reboot fails (consistently).

This is what shows onscreen:

Code: 
BTX loader 1.00 BTX version is 1.02
Console: internal video/keyboard
BIOS drive A: is fd0
... (more drive info)

<dump of registers w/hex values; I can provide it if anyone wants, but can't copy/paste since it's on the physical console>

BTX halted

And that's where I'm at.
Does anyone know what I might try to fix this?

I have searched around for "BTX halted" and found various issues, but no clear solutions that would apply to me...

Maybe I need to rewrite the boot code as indicated in this other post?

I'll try building a live cd/usb stick, so I can investigate the filesystem further.

Some system info:
  • amd64
  • booting from UFS, no ZFS or anything fancy. I do have a hardware raid card, but that seems to initialize fine before it gets to the failure.
  • this is a pretty old system, incrementally updated from 6.x days
 
Interestingly, I made a boot USB stick from FreeBSD-11.4-RELEASE-amd64-memstick.img, booted from that, and got the same failure.
So, that's a clean system that should have the proper boot loader...

Maybe the 11.4 boot loader is incompatible with my system?
I'm going to try making a similar memstick for 11.3 and maybe 12.x and see if it's just 11.4 that's bad.

Also: from researching some, `make installworld` should update the bootcode too. So I imagine I already have the latest, there. This is consistent with the memstick boot failure...
 
I tried making a 12.2 boot stick as well. It also fails to boot, but instead of "BTX halted", I get a screen full of garbage ASCII text.
I think I will be filing a bug report...
 
i speculate that /boot/loader is too large for some setups and it bombs
try to interrupt early boot and change
Default: 0:ad(0,a)/boot/loader
Default: 0:ad(0,a)/boot/loader.old

see if it works
 
covacat said:
> i speculate that /boot/loader is too large for some setups and it bombs
Click to expand...

The system has 8G ram, so I think that shouldn't be the problem.
From discussion on the bug report, it seems like this is a regression that is specific to MBR booting. Haven't tracked down the exact cause yet.

But yes, you are right that previous loaders do work successfully. For instance, I can copy the /boot/loader from a 11.3 memory stick, and it works.

I am curious: what is "loader.old"? I notice that it is significantly smaller than the normal one (224K vs 328K).
Is that not just a copy of the previous OS's loader?
 
jwdevel said:
Also: from researching some, `make installworld` should update the bootcode too.
Click to expand...
It updates the files /boot/loader, /boot/gptboot, /boot/gptzfsboot, etc. It does NOT touch the installed bootloader that's in your freebsd-boot partition. You need to update that yourself ( gpart bootcode -p /boot/gptboot -i 1 ada0 for example).
 
i had the same experience with pxeboot, did not trace the exact problem but older binaries work, newer ones bomb
you can try loader_lua, loader_4th (one of them should already be hard linked to loader)
also you can create a /boot.config to boot /boot/loader.good (11.3?) instead of default
 
You must log in or register to reply here.
Back
Top