On 1/29/2020 11:16 AM, Sasha Levin wrote:
> On Mon, Jan 27, 2020 at 04:51:06PM +0100, Greg Kroah-Hartman wrote:
>> On Sun, Jan 26, 2020 at 08:52:33AM -0500, Sasha Levin wrote:
>>> On Sat, Jan 25, 2020 at 08:03:25PM -0800, Florian Fainelli wrote:
>>> > Hi Greg, Sasha,
>>> >
>>> > Could you backport upstream commit
>>> > de19055564c8f8f9d366f8db3395836da0b2176c ("Documentation: Document
>>> arm64
>>> > kpti control") to the stable 4.9, 4.14 and 4.19 kernels since they all
>>> > support the command line parameter.
>>>
>>> Hey Florian,
>>>
>>> We don't normally take documentation patches into stable trees.
>>
>> Normally we do not, but this is simple enough I've queued it up for 4.19
>> and 4.14. Are you sure it is ok for 4.9? If so, Florian, can you
>> provide a backported version of it?
>
> My objection to taking documentation patches is either that we take all
> of them, or we take none. If we take only select documentation fixes it
> makes a frankenstein Documentation/ directory that might cause more harm
> than benefit.
>
> Let's say I'm looking for netfilter documentation on 4.19, can I trust
> linux-4.19.y or do I look upstream? Right now I know I have to look
> upstream, but if we tell people it's okay to trust the linux-4.19.y docs
> then we might be causing harm to our users when some fixes were
> backported but corresponding documentation fixes weren't.
For a high profile feature/parameter such as kpti it seems to me that
making sure that the documentation reflects what the code supports is a
good way to limit the amount of support requests. For other options, I
would agree with you that back porting them probably makes little sense.
--
Florian
