Where is 2.6.32?
Bill Davidsen
davidsen at tmr.com
Sun Jan 3 05:32:23 UTC 2010
john wendel wrote:
> On 12/31/2009 12:14 PM, Bill Davidsen wrote:
>> Konstantin Svist wrote:
>>> On 12/31/2009 09:10 AM, Bill Davidsen wrote:
>>>> And leaves you with no Fedora patches and the disk performance
>>>> regression issues of 2.6.32. Also a tainted kernel which some
>>>> developers will ignore if you get a trace, etc.
>>>
>>> I thought it's only tainted if there are non-GPL modules compiled in.
>>> For instance, I saw the tainted message whenever I insmod'ed fglrx
>>> driver
>>>
>> You're right, I am assuming he was talking about the nvidia modules
>> which are not GPL, when he mentioned 2.6.32.2+Nvidia. So it would only
>> me tainted if he wanted to have graphics. Or the licensing may have
>> changed, things are not the same for long.
>>
>
>
> Who needs the Fedora patches? I'm not missing them here. Can you tell me
> exactly the patches I'm missing and what they would do for me? If these
> patches are so valuable, why aren't they submitted upstream so the world
> can benefit. Maybe because Linus doesn't want them?
>
> I haven't noticed any disk performance regression/problem. Maybe I don't
> beat it hard enough. hdparm -Tt shows 60.84 MB/s with the fedora kernel
> and 61.09 MB/s with my kernel. I know there's a CFS throughput problem,
> but that's easily fixed.
>
You probably won't see a problem with random access using the hdpart sequential
access test ;-) But there's a thread in LKML something like" 30% regression in
random throughput" and depending on what you do you will really see that.
> My Fedora kernel would also be tainted, since I have to run the Nvidia
> driver in any case.
>
> I don't see any down side to running my own kernel. Plus I save 8MB of
> kernel memory (enough to negate the bloated Nvidia driver), and I enjoy
> the tweaking.
>
I was mentioning 2.6.32 particularly, not building kernels in general. I built
that to try the new video drivers (lock up and run 3x slower than vesa for me),
and BFS (guess I don't trigger anything that shows its advantages, if any). But
I don't bother to build daily kernels unless there a good reason. Been there
done that, ran -ck, -aa, -mm, and -ac kernels, built 2.5 kernels daily, 3-4/day
when CFS patches were coming out a lot, and unless I have a patch to test I
leave it to others.
> Best wishes in the new year!
>
> John
>
--
Bill Davidsen <davidsen at tmr.com>
"We have more to fear from the bungling of the incompetent than from
the machinations of the wicked." - from Slashdot
More information about the users
mailing list