Re: paravirt patches in -mm

[Zachary Amsden <zach@vmware.com>]

Andrew Morton wrote:
> On Sat, 30 Sep 2006 21:46:50 -0700
> Zachary Amsden <zach@vmware.com> wrote:
>
>   
>> Andrew Morton wrote:
>>     
>>> Guys, could you please confirm that these patches from -mm:
>>>
>>> paravirt-remove-read-hazard-from-cow.patch
>>> paravirt-pte-clear-not-present.patch
>>> paravirt-lazy-mmu-mode-hooks.patch
>>> paravirt-combine-flush-accessed-dirty.patch
>>> paravirt-kpte-flush.patch
>>> paravirt-optimize-ptep-establish-for-pae.patch
>>> paravirt-remove-set-pte-atomic.patch
>>> paravirt-pae-compile-fix.patch
>>> paravirt-update-pte-hook.patch
>>>
>>> are suitable for mainline inclusion?
>>>   
>>>       
>> Yes, these are all originally by me, and they are suitable for 
>> mainline.  They are either nops in mainline or are actually enhancements 
>> to mainline, simplifying or optimizing the pte accessors on i386 and 
>> have zero impact (or improved) impact on code generation and 
>> performance.
>>     
>
> OK, thanks.
>
>   
>>  The only regrettable fact is that I broke PAE somewhere in 
>> the middle of the series (thus requiring paravirt-pae-compile-fix as a 
>> separate patch, when I should have rolled the fix into the proper 
>> patch).
>>     
>
> I can find no PAE compile error, and paravirt-pae-compile-fix.patch just
> removes a few config.h inclusions.
>
>   
>>  But if you take the whole batch, there is no problem.
>>     
>
> Well we don't like to break the build partway through a patch series
> because that makes git-bisecting a pain if you happen to land at a bad
> spot.
>   

Yes, I know.

> <ten compiles later>
>
> It all looks good to me. 
>   

Ah, ok, now I know what that was.  It was pre-emptively fixing a later 
PAE breakage caused by some include file manipulations that happen if 
you include pgtable.h in certain assembler files.

Zach