Should we go for Linux 5.15.64 for Core Update 170?

Should we go for Linux 5.15.64 for Core Update 170?

[Peter Müller]

[-- Attachment #1: Type: text/plain, Size: 836 bytes --]

Hello development folks,

earlier today, I came across https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.64, in
which several commits caught my attention, particularly all those in the tcp/netfilter subsystems.

Generally, I am not a fan of updating the kernel in Core Updates while they are already in
testing (unless something is badly broken, of course), since this is an uphill battle, and in
the past has delayed releases quite notably.

Therefore, I would be grateful for input, since I am not too sure how to judge the severity of
these aforementioned kernel commits. Is this something we should bring to our users sooner rather
than later? Does anything strike you as "hey, we have dealt with this bug for ages, and it is
finally resolved upstream"?

Thanks in advance, and best regards,
Peter Müller

Re: Should we go for Linux 5.15.64 for Core Update 170?

[Michael Tremer]

[-- Attachment #1: Type: text/plain, Size: 1568 bytes --]

Hello,

> On 2 Sep 2022, at 17:45, Peter Müller <peter.mueller(a)ipfire.org> wrote:
> 
> Hello development folks,
> 
> earlier today, I came across https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.64, in
> which several commits caught my attention, particularly all those in the tcp/netfilter subsystems.

The kernel log always reads a bit scary. And yes, if in doubt, I would prefer to rather ship one kernel too many.

However, this massively breaks our development model and makes the entire stabilisation phase a lot harder - and we are already struggling with that.

Now, .65 is already out as well. I wouldn’t object to update to this, but I am very short on time this week and really would like to see the update going out soon.

> Generally, I am not a fan of updating the kernel in Core Updates while they are already in
> testing (unless something is badly broken, of course), since this is an uphill battle, and in
> the past has delayed releases quite notably.

Agreed.

> Therefore, I would be grateful for input, since I am not too sure how to judge the severity of
> these aforementioned kernel commits. Is this something we should bring to our users sooner rather
> than later? Does anything strike you as "hey, we have dealt with this bug for ages, and it is
> finally resolved upstream"?

I didn’t spot anything that would be incredibly scary, but I agree that there is an unusually high amount of networking patches in this release.

-Michael

> Thanks in advance, and best regards,
> Peter Müller

Re: Should we go for Linux 5.15.64 for Core Update 170?

[Adolf Belka]

[-- Attachment #1: Type: text/plain, Size: 2146 bytes --]

Hi All,

On 05/09/2022 11:59, Michael Tremer wrote:
> Hello,
>
>> On 2 Sep 2022, at 17:45, Peter Müller <peter.mueller(a)ipfire.org> wrote:
>>
>> Hello development folks,
>>
>> earlier today, I came across https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.64, in
>> which several commits caught my attention, particularly all those in the tcp/netfilter subsystems.
> The kernel log always reads a bit scary. And yes, if in doubt, I would prefer to rather ship one kernel too many.
>
> However, this massively breaks our development model and makes the entire stabilisation phase a lot harder - and we are already struggling with that.
>
> Now, .65 is already out as well. I wouldn’t object to update to this, but I am very short on time this week and really would like to see the update going out soon.
We could release CU170 with its current kernel and do a follow-up CU171 
release with just the new kernel. That way we get CU170 out quickly and 
can fairly quickly and with simple testing get the updated kernel with 
its network patches fixed.

I presume that the networking bugs found with 5.15.59 are also present 
with the kernel that is in CU169 so the impact of issuing CU170 with 
that kernel is not worse than for the existing CU169.

Regards,
Adolf.
>> Generally, I am not a fan of updating the kernel in Core Updates while they are already in
>> testing (unless something is badly broken, of course), since this is an uphill battle, and in
>> the past has delayed releases quite notably.
> Agreed.
>
>> Therefore, I would be grateful for input, since I am not too sure how to judge the severity of
>> these aforementioned kernel commits. Is this something we should bring to our users sooner rather
>> than later? Does anything strike you as "hey, we have dealt with this bug for ages, and it is
>> finally resolved upstream"?
> I didn’t spot anything that would be incredibly scary, but I agree that there is an unusually high amount of networking patches in this release.
>
> -Michael
>
>> Thanks in advance, and best regards,
>> Peter Müller

-- 
Sent from my laptop

Re: Should we go for Linux 5.15.64 for Core Update 170?

[Michael Tremer]

[-- Attachment #1: Type: text/plain, Size: 2401 bytes --]

Hello,

> On 5 Sep 2022, at 11:53, Adolf Belka <adolf.belka(a)ipfire.org> wrote:
> 
> Hi All,
> 
> On 05/09/2022 11:59, Michael Tremer wrote:
>> Hello,
>> 
>>> On 2 Sep 2022, at 17:45, Peter Müller <peter.mueller(a)ipfire.org> wrote:
>>> 
>>> Hello development folks,
>>> 
>>> earlier today, I came across https://cdn.kernel.org/pub/linux/kernel/v5.x/ChangeLog-5.15.64, in
>>> which several commits caught my attention, particularly all those in the tcp/netfilter subsystems.
>> The kernel log always reads a bit scary. And yes, if in doubt, I would prefer to rather ship one kernel too many.
>> 
>> However, this massively breaks our development model and makes the entire stabilisation phase a lot harder - and we are already struggling with that.
>> 
>> Now, .65 is already out as well. I wouldn’t object to update to this, but I am very short on time this week and really would like to see the update going out soon.
> We could release CU170 with its current kernel and do a follow-up CU171 release with just the new kernel. That way we get CU170 out quickly and can fairly quickly and with simple testing get the updated kernel with its network patches fixed.
> 
> I presume that the networking bugs found with 5.15.59 are also present with the kernel that is in CU169 so the impact of issuing CU170 with that kernel is not worse than for the existing CU169.

Indeed. And I am not aware of any reports from users that there are any bugs they are running into.

-Michael

> 
> Regards,
> Adolf.
>>> Generally, I am not a fan of updating the kernel in Core Updates while they are already in
>>> testing (unless something is badly broken, of course), since this is an uphill battle, and in
>>> the past has delayed releases quite notably.
>> Agreed.
>> 
>>> Therefore, I would be grateful for input, since I am not too sure how to judge the severity of
>>> these aforementioned kernel commits. Is this something we should bring to our users sooner rather
>>> than later? Does anything strike you as "hey, we have dealt with this bug for ages, and it is
>>> finally resolved upstream"?
>> I didn’t spot anything that would be incredibly scary, but I agree that there is an unusually high amount of networking patches in this release.
>> 
>> -Michael
>> 
>>> Thanks in advance, and best regards,
>>> Peter Müller
> 
> -- 
> Sent from my laptop