Manual or Slackpkg Kernel Upgrade?

vtel57 · 07-21-2021, 06:09 PM

I'm just curious here...

I've been running Slackware for nearly 20 years as my primary OS. During that time I've almost always used the generic kernels w/ initrd, and manually perform kernel upgrades.

In my most recent installation of Slackware64 - 14.2, I decided to just stick with the default huge kernel. I've gotten lazy, plus I really don't have anything going on these days that requires the initrd method of booting.

So, a kernel upgrade notice just came up today, and I decided to let Slackpkg do the upgrade for me (again, laziness). All went fine and dandy, of course. However, I've always noticed that little comment in /etc/slackpkg/blacklist that says,

Code:

# Automated upgrade of kernel packages aren't a good idea (and you need to
# run "lilo" after upgrade). If you think the same, uncomment the lines
 # below

So, that all being said... how do you folks upgrade your kernels? Automation or manually?

Thanks in advance for your input.

~Eric

hitest · 07-21-2021, 06:36 PM

Quote:

Originally Posted by vtel57

So, that all being said... how do you folks upgrade your kernels? Automation or manually?

In the past I rolled my kernels manually (Eric has a great guide). These days I automatically upgrade my Slackware units (one unit uses LILO and a newer box uses GRUB). The slackpkg utility works well for me.

vtel57 · 07-21-2021, 06:37 PM

I should have actually made this a Poll, huh? Darn... didn't even think about that.

AH! I could add the poll after the fact.

.

Chuck56 · 07-21-2021, 07:20 PM

I selected Other. Scripted kernel install (not upgrade) with multiple kernels available to run.

Spinlock · 07-21-2021, 07:23 PM

/etc/slackpkg/blacklist has had that verbiage changed:

Code:

# Automated upgrade of kernel packages may not be wanted in some situations;
# uncomment the lines below if that fits your circumstances:

I built a fixboot.sh script that I run after each kernel upgrade, which keeps my elilo install sailing along. Heck, I'm about to put it in rc.local.shutdown, which I should have probably done years ago.

Git repo here for those interested

slacker1337 · 07-21-2021, 07:29 PM

I go the slackpkg route as well. I cooked up a script that runs at shutdown that checks the running kernel against the version in /var/log/packages, then it does all the magic.

That said, I only ever had issues with automatic upgrades through slackpkg if I did something stupid.

pghvlaans · 07-21-2021, 10:12 PM

It's a little bit of both for me. I threw together an overall upgrade script that uses sed to comment and uncomment the kernel entries so slackpkg downloads the packages but doesn't upgrade the kernels. It uses installpkg to do the actual installation part.

twy · 07-22-2021, 12:24 AM

I let slackpkg install kernel-firmware and kernel-headers only, no other kernel packages.

I download and install the kernel from kernel.org. I usually run the kernel version that is in slackware, which is now 4.4.276. My .config is from the modular slackware kernel, but I then edit it to include only the devices that I actually have. The process is like...

cd /usr/src
tar xvf ~/downloads/linux-4.4.276.tar.xz
cd linux-4.4.276
cp -a ../linux-4.4.261/.config .
make oldconfig
make -j6 bzImage
make -j6 modules
make modules_install
cp -a arch/x86/boot/bzImage /boot/vmlinuz-4.4.276
cp -a System.map /boot/System.map-4.4.276
cd /boot/
rm System.map
ln -s System.map-4.4.276 System.map
mkinitrd -F -c -k 4.4.276 -s /boot/initrd-tree-4.4.276 -o /boot/initrd.gz-4.4.276

Then, I edit boot entries in /etc/lilo.conf to use the new kernel and keep old kernel as another backup boot option. Finally, run:

lilo

Before rebooting, I uninstall the nvidia driver:

cd <to where nvidia driver is>
./NVIDIA-Linux-x86_64-340.108.run --uninstall

After reboot, reinstall the nvidia driver:

./NVIDIA-Linux-x86_64-340.108.run

I keep copies of /usr/lib/libEGL.la and /usr/lib64/libEGL.la and restore them after the above, because something happens to them.

I use linux device mapper raid (/dev/mapper md-raid), logical volume manager (lvm), and cryptsetup (luks) to setup my root (/) device. Therefore, I have to use initial root disk (initrd) on /dev/ram0 to bring up my actual system root disk (/ = /dev/mapper/cvg1-root) [luks en(c)rypted volume group vg1, logical volumes root and swap]. A raid1 disk is for /boot (raid1\ext4) and a raid6 disk for the rest of / (raid6\luks\lvm\ext4). My /etc/mkinitrd.conf is like this:

# mkinitrd.conf.sample
# See "man mkinitrd.conf" for details on the syntax of this file
#
#SOURCE_TREE="/boot/initrd-tree"
#CLEAR_TREE="0"
#OUTPUT_IMAGE="/boot/initrd.gz"
#KERNEL_VERSION="$(uname -r)"
#KEYMAP="us"
#MODULE_LIST="ext4"
LUKSDEV="/dev/md1"
#LUKSKEY="LABEL=TRAVELSTICK:/keys/alienbob.luks"
ROOTDEV="/dev/cvg1/root"
ROOTFS="ext4"
#RESUMEDEV="/dev/sda2"
RAID="1"
LVM="1"
UDEV="1"
#MODCONF="0"
## load microcode, starting with linux 4.4.110
## to mitigate MELTDOWN and SPECTRE bugs
MICROCODE_ARCH="/boot/intel-ucode.cpio"
#WAIT="1"

I installed the intel microcode package from slackbuilds.org. This is needed to enable the cpu vulnerabilities mitigations (/sys/devices/system/cpu/vulnerabilities/*). The mkinitrd prepends intel-ucode.cpio to the initrd.gz so that the microcode is loaded before anything else happens. Upgrading the microcode package should be done before running mkinitrd.

I use lilo like this:

# LILO global section

append=" vt.default_utf8=1 "
boot = /dev/md0
lba32
raid-extra-boot = mbr-only
compact
message = /boot/boot_message.txt
prompt
timeout = 1200

change-rules
reset

vga = normal

# End LILO global section
# Linux bootable partition config begins

## default kernel
# 4.4.276
image = /boot/vmlinuz-4.4.276
# addappend = " spec_store_bypass_disable=on "
root = /dev/ram0
initrd = /boot/initrd.gz-4.4.276
label = 4.4.276
read-only

## backup kernel
# 4.4.261
image = /boot/vmlinuz-4.4.261
# addappend = " spec_store_bypass_disable=on "
root = /dev/ram0
initrd = /boot/initrd.gz-4.4.261
label = 4.4.261b
read-only

## slackware kernel
# 4.4.276
image = /boot/vmlinuz-4.4.276
# addappend = " spec_store_bypass_disable=on "
root = /dev/ram0
initrd = /boot/initrd.gz-4.4.276
label = 4.4.276s
read-only
# End lilo config

Upgrading the kernel takes some work for me, but it is routine. By doing it this way, setting it up myself, at least I know how it works and can probably fix it if there is a problem. The expected problem is an occasional disk failure.

andrew.46 · 07-22-2021, 12:32 AM

I manually do kernel upgrades and also 'roll my own' for my default kernel. Mind you I am retired and I only run one system so I have time for such things

.

cycojesus · 07-22-2021, 12:55 AM

Automated script (running headless on servers too) to upgrade&co everything + rebuilt initrd at the end of the process. Hasn't failed me in many years now. Always running latest kernel, not even keeping the previous one as a backup, no need. Worst case I'll boot up an install disk and fix from there but I think last failure was hardware and the one before was when still using lilo and kernel-huge got too big. Moved to EFI and kernel-generic.

Didier Spaier · 07-22-2021, 01:23 AM

Neither manual nor slackpkg, the attached scrip called when running "slapt-get --upgrade":

installs the new generic kernel (no huge kernel provided),
removes the not currently running installed kernels if shipped in Slint (not the custom ones),
makes an initrd for each generic kernel which does not have one, including a LUKS key if need be to unlock an encrypted target drive,
updates the grub config file.

brianL · 07-22-2021, 03:05 AM

Always used slackpkg, for every kernel update on 14.2 and current (when I was running it). No problems, easy.

Code:

brian@slackdesk:~$ uname -r
4.4.276

SCerovec · 07-22-2021, 03:54 AM

My wrap up after an new kernel is installed

Once i move to elilo or other boot loaders i might add more to the script

patches are welcome meanwhile (no promises of inclusion before audit tho)

TheRealGrogan · 07-22-2021, 04:41 AM

Manually, the old fashioned way :-)

GazL · 07-22-2021, 04:57 AM

Another 'other' here.

I review *.new files manually, but the rest is handled by scripts, including kernel, initrd, and /boot/efi/ (elilo) updates.