VMware patches for Ubuntu

[Stefan Bader]

Alok Kataria wrote:
> On Wed, 2009-01-14 at 07:38 -0800, Stefan Bader wrote:
>> Hi Alok,
>>
>> thank you for the patches. Unfortunately by now Intrepid moved into maintenance 
>> mode. Meaning we have to go through a stricter mode of requirements. Normally 
>> only major bug fixes or very simple changes. I will have a look at your patches 
>> but I cannot promise we will be able to apply them.
> 
> Hi Stefan,
> 
> As i have mentioned earlier in this thread, ( please see below for a
> snippet from my initial mail)
> 
> -------------------------------------------------------------------------
> These patches fix some of the TSC related bugs which occur only in a
> virtualized environment. The reason they are important for VMware is
> because we rely on TSC as the primary clocksource for timekeeping rather
> than a paravirtualized clocksource.
> 
> So these patches are pretty important for correct timekeeping under
> VMware.

Hi Alok,

I finally got some time for a review. The patchset looks ok to me as well. The 
only downside is that this requires an ABI change. But at least for Intrepid 
there are other changes that will require one. So that would fit. I am going to 
start a formal SRU process.

Stefan
> -------------------------------------------------------------------------
> 
> Apart from the incorrect timekeeping problem, these BUG could result in
> Ubuntu guest hangs under highly overcommitted virtualized environment.
> 
> Given the above facts, IMHO these patches should be considered as "major
> bugfix".
> 
>>From the "simple changes" POV, apart from the hypervisor detection code,
> most of the other changes do not have any affect when run on native
> hardware. IOW most of the code is executed only when you are running
> under a VMware hypervisor. 
> 
> Please let me know if you have any other concerns.
> 
> Thanks,
> Alok
>> Regards,
>> Stefan
>>
>> Alok Kataria wrote:
>>> Hi Tim,
>>>
>>> I was supposed to backport these TSC patches - which are required when
>>> run under VMware - for the intrepid tree, but amidst my vacation and
>>> other things these patches just fell off my radar.
>>> I have backported these patches for the intrepid tree and am attaching
>>> all the seven patches with this mail.
>>> Can you please have a look at these patches and apply them to the
>>> intrepid git tree.
>>>
>>> Thanks,
>>> Alok
>>>
>>> On Fri, 2008-11-07 at 14:18 -0800, Alok Kataria wrote:
>>>> On Fri, 2008-11-07 at 13:33 -0800, Tim Gardner wrote:
>>>>> Alok Kataria wrote:
>>>>>> On Fri, 2008-11-07 at 12:15 -0800, Tim Gardner wrote:
>>>>>>> Alok Kataria wrote:
>>>>>>>> Thats cool, so that would mean you can easily cherry pick these patches
>>>>>>>> for the intrepid tree. Can you let me know for which release can we
>>>>>>>> expect to see these patches.
>>>>>>>>
>>>>>>> If I don't run into issues with Intrepid, then these patches could end
>>>>>>> up in -proposed within the next week or two. 
>>>>>> Great, let me know how it goes. Also once the patches get into proposed
>>>>>> tree i can then have a kernel built off that tree for my internal
>>>>>> testing.
>>>>>>
>>>>>> Thanks,
>>>>>> Alok
>>>>>>
>>>>> I think I'll wait until I see what you come up with for the 2.6.24
>>>>> backported patches. For 2.6.27 the first commit
>>>>> b2bcc7b299f37037b4a78dc1538e5d6508ae8110 wants to patch a non-existing
>>>>> file arch/x86/include/asm/cpufeature.h, so I guess its already off into
>>>>> the weeds.
>>>> Ah...yes all these are failing because the x86 architecture specific
>>>> header files were moved into arch/x86/include...,in 2.6.28-rc1.
>>>>
>>>> That means i will have to backport these for intrepid too, let me do
>>>> that too then.
>>>>
>>>> Thanks,
>>>> Alok
>>>>> rtg
>>
> 


-- 

When all other means of communication fail, try words!