Red Hat's "obfuscated" kernel source [LWN.net]

Red Hat's "obfuscated" kernel source

Posted Feb 28, 2011 18:22 UTC (Mon) by clugstj (subscriber, #4020)
In reply to: Red Hat's "obfuscated" kernel source by BenHutchings
Parent article: Red Hat's "obfuscated" kernel source

Did you ask them why they made this change?

to post comments

Red Hat's "obfuscated" kernel source

Posted Feb 28, 2011 23:33 UTC (Mon) by jerub (guest, #73225) [Link] (9 responses)

Yes, I have asked why they have made this change. I tracked down and interrogated several Red Hat engineers on this issue, and while they were very reticent to speak about it, I discovered the following things:

- It is not about Centos.

- The primary motivation was to make it harder for Oracle Enterprise Linux to repackage the work that Red Hat do.

- The kernel tarball inside the srpm is created from a git tree that is only accessible to Red Hat engineers.

- This change to the way that kernels are dealt with inside Red Hat has angered and frustrated engineers who work on the product. Employees of the company are Not Happy.

- The orders to do this, to make it harder to rebuild the kernel with and without patches, and to make it harder to extract specific patches from the Red Hat kernel came from the top. This is with the knowledge of, and by the order of, the CEO: Jim Whitehurst.

- There is a web interface (somewhere!) that is available that will allow you to specifically omit specific patches and download a new kernel. This is a clunky web front end to the git tree.

- An Oracle engineer I interviewed on this matter greeted this news with in-credulousness, and quickly got out his notebook so he could provide me with links to the various public git trees that oracle maintains of their kernels, and showed me where I could download them from.

Red Hat's "obfuscated" kernel source

Posted Mar 1, 2011 11:20 UTC (Tue) by nix (subscriber, #2304) [Link]

- There is a web interface (somewhere!) that is available that will allow you to specifically omit specific patches and download a new kernel. This is a clunky web front end to the git tree.

Well, that makes the whole thing completely pointless, doesn't it? One bit of automation, a pile of rather slow downloading and bingo, the patches are split out again. Only effect: lots of extra bandwidth cost for RH.

Red Hat's "obfuscated" kernel source

Posted Mar 2, 2011 17:10 UTC (Wed) by jnh (subscriber, #69758) [Link]

> - The primary motivation was to make it harder for Oracle Enterprise
> Linux to repackage the work that Red Hat do.

This makes me ache. For the past several years, I'd been fighting an
on-again-off-again argument with the lead DBA at my (mercifully former)
place of employment about OEL vs. RHEL. We'd run all our Oracle DB and
grid infrastructure systems on RHEL from day 1, and we used RHEL everywhere
we needed support contracts to back up proprietary software installations.
The sysadmins didn't want to add a distro migration to their TODO list,
particularly one with no obvious benefits to the other applications we ran
on RHEL. The DBAs argument was invariably "we probably wouldn't have
problem $foo if we ran OEL, and even if we did Oracle would owe us an
explanation becuase we have a support contract." (This is ignoring that
we obviously had a RHEL contract too, our DBA didn't like to leave his
little world of all-things-Oracle.) During the last argument of this
nature, which was centered around the kernel's swap behavior, I visisted
both Oracle and RedHat's archives, pulled down the kernel patches, and
compared them to see if there was *any* way the OEL kernel would behave
differently. At the time there was only one patch that was even close to
being relevant in the mm subsystem and there was no way it was going to
make a difference in the behavior we were trying to figure out. I used
that information to *quickly* shutdown the argument and bring the focus
off of switching to OEL and back to actually figuring out the root cause.
While I could have still made the comparison by grabbing the patches from
rhn (which I assume is where this web UI lives) it would have been an
inconvenience. Assuming that OEL really is the primary motivation behind
this, I'm embarrassed for RedHat. It's far easier to inconvenience your
customers than it is to inconvenience somebody like Oracle.

Red Hat's "obfuscated" kernel source

Posted Mar 3, 2011 0:05 UTC (Thu) by BenHutchings (subscriber, #37955) [Link] (6 responses)

I wear another hat as a driver developer, and I can say that it is significantly more frustrating to review and test RH's backport of the driver to RHEL 6 than the backport to RHEL 5 - even though the latter is 14 kernel releases older and consequently needs more changes to the upstream driver.

Red Hat's "obfuscated" kernel source

Posted Mar 5, 2011 16:15 UTC (Sat) by BenHutchings (subscriber, #37955) [Link] (1 responses)

Test comment with Iceweasel 4.0; please ignore.

Red Hat's "obfuscated" kernel source

Posted Aug 8, 2012 11:58 UTC (Wed) by xz (guest, #86176) [Link]

Test comment; please ignore.

Red Hat's "obfuscated" kernel source

Posted Mar 5, 2011 16:18 UTC (Sat) by BenHutchings (subscriber, #37955) [Link]

Test comment with Iceweasel 4.0 and add-ons; please ignore.

Red Hat's "obfuscated" kernel source

Posted Mar 5, 2011 16:20 UTC (Sat) by BenHutchings (subscriber, #37955) [Link]

And one final test comment.

Red Hat's "obfuscated" kernel source

Posted Mar 5, 2011 16:36 UTC (Sat) by BenHutchings (subscriber, #37955) [Link]

Sorry, one more test required.

Red Hat's "obfuscated" kernel source

Posted Aug 8, 2012 13:18 UTC (Wed) by nix (subscriber, #2304) [Link]

The question is why this comment of all comments is receiving so many 'test comment' postings :)