Re: EXTERNAL: OT: Linux vendor survey results
Date: Fri, 6 Dec 2013 18:59:34 -0500
Message-ID: <CAHwLALMM+eeujaFOq=H4_-00R90Xp-8QgNrRu6szarcw70OPcA_at_mail.gmail.com>
(I was trying to mention Oracle's public git but then found that initially I couldn't send messages to the list... so I will just add one extra point about the Oracle public git)
Two years ago Redhat started to merge everything into one giant kernel patch before they release it to the public hoping that others won't know what they added and fixed for RHEL customers:
Oracle reverse-engineered redhat's patch and the git repo is available at: https://oss.oracle.com/git/?p=redpatch.git;a=summary
I've worked with guys at Oracle who develop Oracle Linux, and Oracle does have the engineering capability to fix issues -- in fact, besides the UEK & the RHEL kernel, Oracle also offers a patched RHEL kernel that has patches not in RHEL.
Rayson
Open Grid Scheduler - The Official Open Source Grid Engine http://gridscheduler.sourceforge.net/
http://gridscheduler.sourceforge.net/GridEngine/GridEngineCloud.html
On Fri, Dec 6, 2013 at 4:51 PM, Marc Fielding <fielding_at_pythian.com> wrote:
> Don,
>
> Taking a peek at Oracle's UEK source git viewer, the full IB stack seems to
> be there, including change-level detail. See
> https://oss.oracle.com/git/?p=linux-uek3-3.8.git;a=tree;f=drivers/infiniband/core
> for example. According to the UEK3 announcement
> (https://blogs.oracle.com/linux/entry/announcing_the_unbreakable_enterprise_kernel)
> it's just the OFED 2.0 stack. It's dual GPL/BSD licensed, so I don't see a
> reason why Red Hat couldn't incorporate the bits and pieces into their own
> kernel should they choose to.
>
> Cheers,
>
> Marc
>
>
> On Fri, Dec 6, 2013 at 12:04 PM, Don Seiler <don_at_seiler.us> wrote:
>>
>> It's accurate, and I asked them repeatedly about the nature of the bug and
>> the fix and they wouldn't tell me. I'm drafting a blog post about it but was
>> hoping to provide more detail for the reader other than "Oracle said so". My
>> take was that Oracle said they re-wrote the module from scratch, and so
>> wasn't bound by any GPL-type of license forcing them to make their changes
>> available.
>>
>> It was disappointing to say the least, and so long story short we had to
>> abandon Infiniband in favor of 10gbE.
>>
>> Don.
>>
>>
>> On Fri, Dec 6, 2013 at 12:31 PM, Matthew Zito <matt_at_crackpotideas.com>
>> wrote:
>>>
>>> If that's accurate, that's an extremely hostile statement on the part of
>>> Oracle. Linux is open source, and in general, other than some video drivers
>>> that are still a sensitive subject, kernel modules are not closed source.
>>> For Oracle to take all of the work that Red Hat put into building RHEL, copy
>>> all of the bits they *did not* make closed source, and then reimplement just
>>> the pieces they choose and call it proprietary is very troubling.
>>>
>>> Matt
>>>
>>>
>>> On Fri, Dec 6, 2013 at 1:22 PM, Don Seiler <don_at_seiler.us> wrote:
>>>>
>>>> I can also say that OEL/UEK contains some fixes that aren't available in
>>>> RHEL. In my case it was that Oracle UEK had a complete re-write of the OFA
>>>> module (open fabric). The RHEL version of this module causes DirectNFS to
>>>> choke on Infiniband. Oracle is not sharing their re-write back to RH since
>>>> they say it's proprietary. However we're still on RHEL so we had to abandon
>>>> our Infiniband plans.
>>>>
>>>> Don.
>>>>
>>>>
>>
>>
>>
>> --
>> Don Seiler
>> http://www.seiler.us
>
>
>
>
> --
> Marc Fielding
> Senior Consultant
> Pythian - love your data
> fielding_at_pythian.com | twitter.com/mfild | linkedin.com/in/mfielding
>
> --
>
>
>
-- http://www.freelists.org/webpage/oracle-lReceived on Sat Dec 07 2013 - 00:59:34 CET