Re: From which kernel should I upgrade my installed Debian to linux-image-6.1.0-15-amd64?
Hey Stella, hey all:
Am 11.12.23 um 19:38 schrieb Stella Ashburne:
> Thank goodness it only happens once in a blue moon.
May I please clarify some basic (mis-)conceptions?
There's "linux-image-amd64". This is a metapackage that contains nothing
but a constructed dependency upon the latest kernel package, such as now
"linux-image-6.1.0-15-amd64". If any debian package A has a dependency
upon B, then installing A will make apt install B as well. If you then
try to uninstall B, either dpkg will refuse, or apt will offer to
uninstall A beforehand. That's the single one purpose metapackages exist
for: to make sure you have another package installed.[1]
Why does debian provide the metapackage "linux-image-amd64"? That's
because every now and then your real kernel package, such as
"linux-image-6.1.0-12-amd64" should be updated to
"linux-image-6.1.0-13-amd64", or so on. Those are not two versions of
the same linux package, but rather they're two seperate packages that
may be installed simultaneously, which apt itself will not "update" from
one to the other. The best way for debian to ensure everyone has current
kernel versions, is to provide a metapackage with a name that doesn't
change, such as "linux-image-amd64". That metapackage, when updated, apt
will recognize as depending upon (requiring) a previously not installed
package, such as "linux-image-6.1.0-13-amd64"
Metapackage "linux-image-amd64" version...
6.1.52-1 depends on linux-image-6.1.0-12-amd64
6.1.55-1 depends on linux-image-6.1.0-13-amd64
6.1.64-1 depends on linux-image-6.1.0-14-amd64 (nasty!)
6.1.66-1 depends on linux-image-6.1.0-15-amd64
Also there are similar "dependencies" in place and in sync with the
kernel one, such as another metapackage "linux-headers-amd64".
Also there might be other mechnisms (autoremove) in place for the
purpose of uninstalling your outdated kernels for disk space, but they
spare the current one as well as its one predecessor as your backup boot
option, just in case any update might go bad. (blue moon) Everything
will be fine. (fingers crossed)
If you happened to update bookworm last weekend, you might have caught
"linux-image-amd64" in version 6.1.64-1. That would have intentionally
pulled and installed the accidentally very dangerous
"linux-image-6.1.0-14-amd64". Let's call it "nasty". Please make very
sure to never boot that. It is known to possibly fry your ext4,
potentially destroying all your data. That's why debian interrupted the
12.3 point-release process very quickly. Unprecedented? I'm certain
there're plenty blog posts and maybe official statements about that to
come, about this one-in-a-IDK-glitch. (Andy had suggested a decade, that
seems about correct.) Could anyone have installed, or even booted nasty
6.1.0-14 in the meantime? Yes. I have, for instance. By chance it didn't
hurt me this time, but I'm not betting again.
Re your situation: To make sure, I highly recommend uninstalling nasty
"linux-image-6.1.0-14-amd64", provided that you have another functioning
kernel version. But if your metapackage "linux-image-amd64" is still
version 6.1.64-1, that would get uninstalled as well. Result: You'd no
longer receive automatic kernel updates, very bad. So maybe better first
update the entire debian, including all the metapackages, and as of now
you'll get linux-image-amd64 in version 6.1.66-1. That no longer
"depends" upon nasty linux-image-6.1.0-14-amd64, but rathermore -15.
(Wait, please read on first!)
That's what might happen once in a blue moon at the very most, and hurt
noone, hopefully. And then debian rectified it in the middle of the
point-release process, by hastily cancelling release 12.3 and replacing
it with 12.4. New kernel version is now 6.1.0-15, no more ext4 frying.
What could possibly go wrong by hastily releasing a debian kernel?
()brace for impact()
6.1.0-15 brought not only the ext4-bugfix, but along with it introduced
a terrible new bug: Most computers work fine with -15, except for some
of those that have wifi, depending upon the driver. There was a certain
change in Linux's cfg80211 kernel module, which controls wifi. This very
change was adopted in debian's hastily released 6.1.0-15. Whichever
computer is affected, then not only loses wifi, but becomes virtually
unusable, unable to perform simple tasks, such as even to properly shut
down. So -15 is useless for a number of users. Expect blogposts about
that as well.
An unusable kernel (-15) once in a blue moon is no big deal, since you
can always boot back into... the... previous... one... Oh, wait. Can you
see where this is getting to? Oh no, not -14? So if you're fortunate
enough for -15 to be working for you, lucky you, maybe just keep -13 as
a backup. I happen to be among a few users who need to run -13 and keep
-12 as backup, just to be careful. Always have one extra, really. This
weekend that little extra carefulness paid for me. Another time, it will
for you.
Given all the above, in conjunction with autoremove, this might have
made some people's computers stranded with only -14 and -15 to boot
from. Out of which one might toast all their data, and the other one
might go nuts if their computer has wifi, and by chance they're using a
driver that cfg80211 can't cope with. Those users would be then out of luck.
What a terrible mess. /o\
I strongly believe in debian's honesty, because debian is based upon its
social contract.[2] Have you read and understood it? (It's a very
pleasant read.) The debian project is designed to be free and open, and
to be incapable of covering up anything that might have gone wrong. Much
better than that: Debian will publish just that in most detail. Having
screwed up twice in a row, debian will confess, big time. Had I
mentioned blog posts? (We might need a word with the moon regarding its
blue occurrence. Maybe don't do that twice in a weekend, please?) Debian
is empowering users. That includes that debian will never accept
liability for making anyone's computer unusable. If that ever happens,
then it's the user who's done it. Debian focuses on providing a free
operating system to anyone, enabling them to make best use of their
computers. And debian goes great lengths to help users prevent screwing up.
In order to get things straight (get kernels working), debian's
Salvatore Bonaccorso has been wonderful in pinpointing the -15's bug to
the cfg80211 module, and developing a remedy, which will hopefully be
introduced with future 6.1.0-16 maybe, soon to come. Not again hastily:
It's ready when it's ready. That's how debian's been excellent for 30 years.
So please feel empowered, use debian as much as you like, and if you
feel gratitude or appreciation, please feel free to express any of that.
Maybe towards the release or kernel teams, who've been working unpaid
nightshifts for 12.3, which suddenly became 12.4, and who still are, for
the folks who need to solve that -15 wifi bug. What are they doing it
for? For the good, expecting nothing in return. Just enjoy!
(Since you'd explicitly asked for advice, I hope there was not too much
mansplaining.)
[1] (installed. or in other cases, not installed, or in a specific
version, etc.)
[2] https://www.debian.org/social_contract
--
Kevin Price
Reply to: