Rich Freeman wrote:
> On Fri, Jan 3, 2020 at 5:57 AM Dale <[email protected]> wrote:
>> Can you post a ls -al /boot for both kernels and images?  That way I can
>> see how it names them when doing it your way.  If I can make sense of
>> it, I may try doing it that way.  Thing is, it'll change eventually
>> too.  lol
> I use the standard kernel names:
>
> config-4.19.92
> initramfs-4.19.92.img
> System.map-4.19.92
> vmlinuz-4.19.92
> /lib/modules/4.19.92
>
> I create the initramfs using:
> dracut "" 4.19.92
>
> Dracut is going to need the path to the modules more than anything
> else, so I suspect it will work if you substitute 4.19.92 with
> whatever the path of your modules directory is, within /lib/modules.
>
> Also, could you actually post the command lines you're using?  You
> posted 4 fairly long emails elaborating on how everything isn't
> working right, and I don't think you actually posted a single dracut
> command line.  When something isn't working right it is usually best
> to start with what you're actually doing, along with what is happening
> and what you expected to happen.  You mainly covered the last bit of
> those three but left out most of the first two.
>
> I actually use a script to do my kernel updates - this is intended
> mainly for bumps and isn't entirely suitable when I need to change
> things, in which case I usually just build manually following the same
> steps:
> #!/bin/bash
> cd /usr/src/linux || exit
> git pull || exit
> rm -rf /var/tmp/linux || exit
> export KBUILD_OUTPUT=/var/tmp/linux
> make O=/var/tmp/linux oldconfig || exit
> nice -n20 make O=/var/tmp/linux -j12 -l20 || exit
> make O=/var/tmp/linux modules_install || exit
> make O=/var/tmp/linux install || exit
> emerge @module-rebuild || exit
> NEWVER=$(make --no-print-directory kernelversion) || exit
> dracut "" $NEWVER || exit
> grub-mkconfig -o /boot/grub/grub.cfg
>
> (This does all the building in /var/tmp and leaves me with a clean
> kernel source directory.  That is actually the upstream-recommended
> way but it does create the issue that if any package that builds
> kernel modules gets updated it will fail.  I usually just delay
> updating these packages until I do my next kernel update, but I can
> just run this script again to re-create /var/tmp/linux with the
> necessary files to build further modules.  Note that you need a few GB
> in /var/tmp for this to work, and this script doesn't clean up - I
> usually want that directory left for any module updating, and it gets
> cleared on reboot anyway which usually follows a kernel update.  This
> works great on tmpfs if you have the space.
>
> Note also that I'm using upstream stable vanilla sources - I checkout
> a longterm branch which is what is getting pulled at the start.  This
> should work with gentoo sources as well if you just tweak the start.
> I like to maintain more control over what kernel I'm following as I
> tend to use out-of-tree modules like zfs, or experimental ones like
> btrfs, or newer CPUs like Ryzen - for one reason or another just
> following random stable releases is problematic.)
>


Those names make sense but I wonder if I could add sequence numbers on
the end.  Most of the time, -1 works since I use oldconfig a lot.  On
occasion tho, I'll have a -2, like this time, or even a -3.  I don't get
that far as often as I used to tho.

The reason I didn't include a command that was tried, I had so many of
them that I tried and they were spread over several different tabs in
konsole.  I tried changing names of kernels, including locations and no
telling what else.  If I posted them, even I wouldn't be able to make
much sense of it. I'm sure no one else could if I couldn't.  I fiddled
with that for hours.  I don't like going back to a older version since
eventually it will be gone but I did in this case.  Still, I need to
figure out the new one since I will have to use it later on. 

I use gentoo-sources.  I update sometimes but given that I don't reboot
much, it may take a long while to test a kernel.  I've had times where I
build a kernel but later do a newer one and end up never using other
updated kernels.  I use uprecords to tell me what kernels I've used and
for how long. I keep two or three known and well tested kernels around
just in case.  Others I delete if /boot starts taking up to much space. 
I start by removing older kernels that have never been used, then older
kernels and work my way up until I'm left with three or so.  I also make
sure my video drivers will work with updated kernels as well.  Sometimes
the newest ones don't.  I have older cards as a general rule.

I'm going to play with this some more another day.  Health issues has me
staying off this thing a bit. 

Dale

:-)  :-)

Reply via email to