[SRU][F:linux-kvm][PULL] UBUNTU: [Config] kvm: Include master annotations
Roxana Nicolescu
roxana.nicolescu at canonical.com
Thu Oct 26 13:48:16 UTC 2023
On 23/10/2023 18:05, Stefan Bader wrote:
> On 23.10.23 16:47, Thibault Ferrante wrote:
>> BugLink: https://bugs.launchpad.net/bugs/2039930
>>
>> SRU Justification:
>>
>> [Impact]
>>
>> Include master annotations and remove disabling overwrites not
>> related to hardware support.
>> This will ease support for kvm derivatives until EOL.
>
> This is probably not what should be done. This kernel differs a lot in
> what is enabled not by accident but intentionally. However it is not
> ideally named. Unlike that name suggest this was not created as a VM
> kernel but as a fast-boot minimal kernel to quickly deploy and bring
> up. So it has not only lots of hardware disabled but also just a
> minimal support for other things to keep the footprint as small as
> possible.
>
> Unfortunately, due to the naming, it got used for things we did not
> intend but even then we would only enable selected things if we really
> had to. And IMO we should stick to that.
>
> -Stefan
>
>>
>> [Fix]
>>
>> As this a stable kernel, this may be too much already modifications,
>> but those are limited
>> to enable features which are already enabled in master and shouldn't
>> break existing application.
>> Only one option was disabled, CONFIG_EMBEDDED, which was unused.
>>
>> [Test Plan]
>>
>> Built and booted.
>> Boot time are similar, the size of the generated kernel is increased
>> by a few percents.
>> Specific look should be given while doing the next cranking for
>> regression.
>>
>> [Where problems could occur]
>>
>> Quite a number of configuration has been changed.
>> Those are mainly focused on protocol/network, kernel debugging
>> features (BPF/Kprobes/Ftrace),
>> or filesystems support, including NFS.
>>
>> [Other Info]
>>
>> If this change land without issues, we may want to port these changes
>> to other kvm derivatives.
>>
>>
>> ----------------------------------------------------------------
>> The following changes since commit
>> dad75ffb821b7052f17c8e0c44e96a0d6ef2b21f:
>>
>> UBUNTU: Ubuntu-kvm-5.4.0-1102.108
>>
>> are available in the Git repository at:
>>
>> git+ssh://thibf@git.launchpad.net/~thibf/+git/focal-linux lp_2039930
>>
>> for you to fetch changes up to 5c470107ed22c965234635b77bcc427fbcd46135:
>>
>> UBUNTU: [Config] kvm: Match master misc configs
>>
>> ----------------------------------------------------------------
>> Thibault Ferrante (7):
>> UBUNTU: [Config] kvm: Include master annotations
>> UBUNTU: [Config] kvm: Enable kernel debugging options
>> UBUNTU: [Config] kvm: Enable kernel network/protocol options
>> UBUNTU: [Config] kvm: Enable kernel filesystems options
>> UBUNTU: [Config] kvm: match NLS_CODEPAGE config
>> UBUNTU: [Config] kvm: Enable kernel scheduler options
>> UBUNTU: [Config] kvm: Match master misc configs
>>
>> .../abi/5.4.0-1101.107/amd64/kvm.modules | 19 -
>> debian.kvm/config/annotations | 7711 +++++++++++++----
>> 2 files changed, 5833 insertions(+), 1897 deletions(-)
>>
>>
>
I am also inclined to say we should not do this. I think it's perfectly
normal to have kernels that don't match the main one. But the problem
here at least is that these
changes are not documented.
Did you talk to someone or have some documentation backing the config
changes you made? The kvm kernels have been a bit of mistery to me,
apart from the fact they
are supposed to be "minimal".
As a side note, I am not sure how much it will improve the maintenance
of these kernels to be honest. Yes, I know many tests fail
because of this, but at least for adt some are either solved or hinted.
Roxana
More information about the kernel-team
mailing list