Re: linux-next: strange commit in the parisc-hd tree

From: Helge Deller
Date: Tue Jan 12 2016 - 16:44:11 EST


On 11.01.2016 23:39, Michael Ellerman wrote:
> On Mon, 2016-01-11 at 22:46 +0100, Helge Deller wrote:
>> On 10.01.2016 23:28, Stephen Rothwell wrote:
>>> I noticed the following commit in the parisc-hd tree today:
>>>
>>> 43122681dd75 ("Fix __ARCH_SI_PREAMBLE_SIZE on parisc")
>>>
>>> which among other things says:
>>>
>>> This patch is not planned to be committed as it currently is.
>>
>> Right, I split it up into two.
>>
>>> [...] Please remove it from linux-next inclusion until it is ready.
>>
>> Sure, I just removed it now.
>>
>>> Delibeately introducing a patch to linux-next that may well cause build breakage
>>> in other architectures just before the merge window opens is a bit antisocial.
>>
>> That was in no way my intention.
>> In contrary, I wanted to see if other platforms are affected by the same bug without knowing.
>> Do you by any chance know if it broke some of the existing platforms?
>> If yes, that would be valuable info...
>>
>> Just in case people want to know what this is all about:
>> On parisc I found __ARCH_SI_PREAMBLE_SIZE to have a wrong value in 64bit kernel builds.
>> Patch which fixes it is here:
>> https://patchwork.kernel.org/patch/7994821/
>>
>> To prevent other (existing and future) architectures to run into the same
>> problem I wanted to test this patch via linux-next and inform them:
>> http://git.kernel.org/cgit/linux/kernel/git/deller/parisc-linux.git/commit/?h=test__ARCH_SI_PREAMBLE_SIZE
>
> That doesn't fire on powerpc, ie. we have it correct.

It really seems that parisc was the only architecture left which had a wrong value for __ARCH_SI_PREAMBLE_SIZE.
I followed Stephens suggestion to try it via the zero day build farm and it was only reported one problem (for parisc).
Tree is here: http://git.kernel.org/cgit/linux/kernel/git/deller/parisc-linux.git/log/?h=for-next-test

> Though I wonder why it's not just defined using offsetof? Possibly just that
> it's awkward with the way the headers are organised.

Yes.

Helge