On 2015-12-22 20:04 -0500, Phillip Susi wrote:
> On 12/22/2015 02:47 PM, Sven Joachim wrote:
>> Both the original report ("Kernel: Linux 3.16.0-4-amd64 (SMP w/4
>> CPU cores)") and the followup in
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=808380#25 show
>> that Jacob had already running 3.16.0-4. If he had been using
>> 3.16.0-3, the problem would not have shown up since modprobe does
>> not try to load 3.16.0-4 modules into a 3.16.0-3 kernel.
>
> That's what I've been saying.
Except that you insisted that the new kernel and its modules would _not_
replace the old one. But hopefully this is clear now.
>> Debian does not do this, to avoid uncontrolled proliferation of
>> kernels. In fact, the 3.16 kernel has been at the 3.16.0-4 ABI for
>> over a year (the last ABI bump was in November 2014, version
>> 3.16.7-1).
>
> If that is the case, then they screwed up and broke the ABI without
> bumping its revision and the bug should be reopened and reassigned to
> linux so they know they broke the ABI and need to bump the revision.
The ABI of the kernel is only bumped if third-party modules would be
broken and need to be rebuilt. Added symbols don't result in an ABI
bump, if that breaks loading in-tree modules (as is the case here), the
user is expected to remedy the situation by rebooting. Which is
necessary to complete a kernel upgrade anyway.
If you don't like this decision, take it up with the Debian kernel
team.
Cheers,
Sven
