| |||||||||||||
Oracle's "Unbreakable Enterprise Kernel"Oracle's "Unbreakable Enterprise Kernel"Posted Sep 20, 2010 17:01 UTC (Mon) by masoncl (subscriber, #47138)Parent article: Oracle's "Unbreakable Enterprise Kernel"
Hi everyone,
One of the goals of this kernel was to stay as close to 2.6.32.stable as we could. The sources are here in git, they won't be rebased:
http://oss.oracle.com/git/?p=linux-2.6-unbreakable.git;a=...
The main differences from mainline:
*) semtimedop optimizations. I posted these to the list a while ago, and Manfred took things in a less complex direction. He was waiting for me to fully benchmark the less complex version, but we ran out of time in the release cycle and had to focus on other things. Oracle hammers on the IPC lock, so these made a big difference, and now I finally have time to properly benchmark his approach against mine.
*) Ocfs2
There are other bits and pieces, but we resisted the urge to pile things in.
The solid state disk access number came on a huge machine, and the improvements came from getting rid a lock in the driver and enabling it for softirq affinity code without taking any of the request locks.
Over the next 12 months we'll be getting an update prepared to a new mainline version, and trying to hammer on upstream kernels as much as we can to reduce our patch count even more.
-chris
(Log in to post comments)
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 17:20 UTC (Mon) by jetsaredim (guest, #70230) [Link]
All,
I'm one of Chris' co-workers.
The associated rpms are freely available to all on the Oracle Public Yum server (http://public-yum.oracle.com). These repos mirror the content available on the distribution ISOs, but not bugfixes or errata. We even provide repo files to your convenience.
Specifically, the src rpm for the new kernel is located at http://public-yum.oracle.com/repo/OracleLinux/OL5/5/base/....
Enjoy. :)
Jared
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 17:43 UTC (Mon) by daney (subscriber, #24551) [Link]
A perfectly good flame-fest, and someone has to come along and ruin it with facts.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 19:41 UTC (Mon) by masoncl (subscriber, #47138) [Link]
Well, after things were basically working, we started fixing all the real bugs reported by an updated checkpatch.pl
Any lines that obviously should have been longer than 80 chars were joined back to their proper length, and curly braces were moved to their proper locations. Tabs were all changed to 3 spaces each, we couldn't quite decide on 2 vs 4.
After that was done, we forward ported devfs. Proper disc management code in Linux was long overdue.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 20:06 UTC (Mon) by ummmwhat (guest, #54087) [Link]
:) LOL
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 23, 2010 16:53 UTC (Thu) by walex (subscriber, #69836) [Link]
Nothing to laugh about 'devfs', it still has some crucial advantages over the abomination called 'udev', and as to that, please consider that 'udev' actually relies on something that is equivalent (but badly) to 'devfs', please check the output of 'cat /sys/block/sda/dev' for a laugh on how clever and nasty GKH is.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 23, 2010 21:32 UTC (Thu) by zlynx (subscriber, #2285) [Link]
The device numbers are critical information for creating a device node. It isn't equivalent to devfs at all.
Where devfs got into trouble was the creation and destruction of device nodes and the permissions on the nodes.
There is hardly any advantage to devfs over udev when a daemon has to be running to update device node permissions according to system policy. If you have that daemon it may as well be responsible for creating the nodes too.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 23, 2010 22:31 UTC (Thu) by njs (guest, #40338) [Link]
A picture of Greg with his goggles pushed up! You're right, that is clever.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 23, 2010 23:16 UTC (Thu) by bronson (subscriber, #4806) [Link]
That's quite a look of surprise on his face!
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 18:34 UTC (Mon) by ummmwhat (guest, #54087) [Link]
Thanks!
This is what I asked for!
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 18:46 UTC (Mon) by NightMonkey (subscriber, #23051) [Link]
FYI, you've got some 404'd links on the public-yum page:
Oracle Linux -> http://www.oracle.com/technologies/linux/technical.html -> 404
Also, that download page ( http://edelivery.oracle.com/EPD/GetUserInfo/get_form?call... ) has some seriously restrictive language which must be agreed to before downloading. Is that actually compatible with the GPL v2 to have all those additional restrictions attached before downloading the software?
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 19:15 UTC (Mon) by michaeljt (subscriber, #39183) [Link]
> Also, that download page ( http://edelivery.oracle.com/EPD/GetUserInfo/get_form?call... ) has some seriously restrictive language which must be agreed to before downloading. Is that actually compatible with the GPL v2 to have all those additional restrictions attached before downloading the software?
For Oracle Linux I see two restrictions - one that the person downloading must accept the GPLv2, and the other is the export restriction stuff that they are required to put there according to US law. I think that if the GPL doesn't allow that then it forbids any US company from distributing the Linux kernel altogether unless they rip out at least the cryptography bits.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 20, 2010 20:21 UTC (Mon) by nix (subscriber, #2304) [Link]
> the person downloading must accept the GPLv2
Only if they plan to redistribute it themselves (in which case the act of redistribution itself constitutes acceptance).
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 21, 2010 20:09 UTC (Tue) by butlerm (subscriber, #13312) [Link]
"in which case the act of redistribution itself constitutes acceptance"
Strictly speaking that is not true. No one "accepts" an ordinary license. The liability is entirely in terms of getting sued if you go beyond the rights granted to you by the license itself (or the other rights you have license or no license).
Of course this wanders into the bizarre world of shrinkwrap and other post-sale/post-delivery "licenses", and how and why they actually do what they claim to. At least there is substantial and identifiable consideration in most cases there though, making the arrangement look much more like a contract than a blanket grant to whoever and whomever.
Oracle's "Unbreakable Enterprise Kernel" Posted Sep 22, 2010 12:23 UTC (Wed) by nix (subscriber, #2304) [Link]
Yeah, you win the pedantry award. You don't accept a license: you choose to exercise the rights it grants you.
|
|||||||||||||