Lee Damon via freebsd-stable wrote on 2019/03/01 22:53:
After discussion with Bob Bishop (thanks for the help!) I've tried to do
the following to upgrade one of the old boxes I mentioned previously.
cd /usr/src
tar ... .
rm -rf .??* *
svn checkout httpg://svn.freebsd.org/base/releng/10.3 /usr/src
compile, installkernel, installworld...
Now that the host is running RELENG the next step was to update from
10.4 to 11.2 via freebsd-update
freebsd-update
freebsd-install
freebsd-update upgrade -r 11.2-RELEASE
freebsd-update install
so far, so good. Now it all falls apart
shutdown -r now
... why isn't the host coming back? Oh look, kernel panic.
Fatal trap 12: page fault while in kernel mode
cpuid = 1; apci id = 01
fault virtual address = 0x84
fault code = supervisor read data, page not present
I went back from freebsd-update to source upgrades few years ago and now
use exclusively source builds (build it on powerful build machine and
distribute it to clients thru NFS so clients can just run make
installkernel and make installworld) because I was bitten by failed
freebsd-update upgrade many times...
Google searches find references to the same panic type in VMs running
11.1, including https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=220923
The differences are, that's 11.1 not 11.2 (I would presume the fix made
it into 11.2 but maybe not) and most notably, that's against VMs and the
host I'm doing this on is bare iron (Sun x4500).
Still, I gave the two entries in /boot/loader.conf a try, no joy.
Exactly the same panic. Recording the boot with slow-mo shows the panic
happening just after the USB devices are enumerated by the kernel. It
never even tries to mount root.
I am able to boot to kernel.old, which appears to be my old 10.4-STABLE
kernel. So now I'm kind of stuck. The update has already modified the
config files as part of the first pass so rolling back may be a problem
and moving forward seems unwise.
I have only one x4500 but I have three x4540s running 11.2-STABLE (also
installed from source) just fine.
Anyone have any brilliant suggestions? I'm thinking of trying to compile
11.2-RELENG in /usr/src so I can try installing that kernel but that'll
take several hours at least (it's an old box).
If you can boot with the old 10.4 kernel and go online, just fetch
kernel.txz from the net:
http://ftp.freebsd.org/pub/FreeBSD/releases/amd64/11.2-RELEASE/kernel.txz
and unpack it to /boot/kernel112 then you can try to reboot a manually
select to boot this kernel instead of default /boot/kernel.
If you cannot access the boot loader prompt you can try "nextboot" command.
1) unpack the kernel
2) set nextboot: nextboot -k kernel112
3) shutdown -r now and hope for a luck
If your machine boots fine with 11.2 kernel, you can fetch sources and
rebuild kernel and userland for 11.2 as usual.
Or you can try to fetch and unpack base.txz
http://ftp.freebsd.org/pub/FreeBSD/releases/amd64/11.2-RELEASE/base.txz
over your current files. It can make a mess but you can always clean it
with "make delete-old & make delete-old-libs"
Miroslav Lachman
_______________________________________________
freebsd-stable@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-stable
To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
