On 2017-06-20 I did a 'dnf update' on my server. The update threw
some errors and
when I did a 'reboot' with the new kernel
(vmlinuz-4.11.5-200.fc25.x86_64) it failed without
giving me a command prompt. So, I booted my previous kernel
(4.10.17-200.fc25.x86_64).
I got a login prompt and logged in. Most commands crashed and sent
be back to the
login prompt. I was able to look in /var/log/messages and find:
Jun 20 09:16:10 yoda kernel: traps:
cpp[2655] trap invalid opcode ip:7f08a64762a4 sp:7ffd14da8df0
error:0
Jun 20 09:16:10 yoda kernel: in
libc-2.24.so[7f08a6443000+1bd000]
and:
Jun 20 09:16:28 yoda kernel: traps:
groupadd[2707] trap invalid opcode ip:7fe7cf40b2a4
sp:7ffdca8072f0 error:0
Jun 20 09:16:28 yoda kernel: in
libc-2.24.so[7fe7cf3d8000+1bd000]
lots of the above errors. I couldn't even start the
network.
Here is what I did to fix it:
# USB stick is Fedora 25 Server install
#
# unplug 2nd drive of mirror then boot usb stick
# at first install screen, get to console command prompt:
# (alt is the LEFT alt key)
ctrl+alt+F2 <-- might be F3, don't remember
# assemble raid with only one disk, this is /
mdadm -A -R /dev/md2 /dev/sda2
mkdir /mnt/ROOT
mount /dev/md2 /mnt/ROOT
# symlink to repros directory
ln -s /mnt/ROOT/etc/yum.repos.d /etc/yum.repos.d
dnf --config=/mnt/ROOT/etc/dnf/dnf.conf
--installroot=/mnt/ROOT reinstall glibc\* cpp\*
#reboot without usb stick
#if above reboot works, reboot with both raid drives and
sync them:
mdadm /dev/md2 --re-add /dev/sdb2
If you don't have RAID, skip those parts. You'll need to know which
partition is your /
I don't have a /boot partition so there was no need to mount that.
I broke the raid mirror so that if this didn't work, I would have a
known place to go
back to and try something else.
All is well now. I'm just putting this out there in hope that it
will help someone.
Bill
PS. Note that the newest kernel (vmlinuz-4.11.5-200.fc25.x86_64)
still does not boot
to a login or command prompt.