Con Kolivas [interview] has released O17int against the 2.6.0-test3 Linux development kernel with the following comment:
"This patch prevents any single task from being able to starve the runqueue that it's on. This minimises the impact a poorly behaved application or a malicious one has on the rest of the system. If an interactive task (eg blender) spins on wait on a cpu hog (X), other tasks will be less affected by priority inversion occurring between X and blender (ie X will still suffer but the machine wont come to a standstill). Broken apps will improve at least partially with this, but they should be fixed for good performance."
His full explanation follows.
Nick Piggin (me) has submitted a new process policy scheduler for testing on linux-kernel. See my post here for the announcement.
A quick rationale why. The process (or CPU) scheduler is one of the most noticeable components of the kernel to a desktop user, and a crucial to any workload. This is one area that has had constant complaints during 2.5 development, albeit only from desktop users and only a vocal minority. I would hate to see a user upgrading to 2.6 having a worse experience.
Con Kolivas [interview] has released O16.2. The resulting thread is lengthy and quite informative, discussing what works, what doesn't, and why...
. More CPU scheduler changes
. The regression with reaim which was due to the CPU scheduler changes
seems to have largely gone away, but it was never a large effect in my
testing. Needs retesting please.
. A series of Cardbus driver updates.
The full changlog follows.
Alan Cox [interview] has released 2.4.22-rc2-ac3.
Complete changelog within.
Andrew Morton [interview] release 2.6.0-test3-mm2 today. He notes:
. Zillions of bugfixes . Updated framebuffer drivers . More CPU scheduler tweaking.
Con Colivas [interview] has released his latest interactivity patch, O15int. He notes that this version of the patch "addresses the problem of tasks that preempt their children when they're forking, wasting cpu cycles until they get demoted to a priority where they no longer preempt their child." He goes on to summarize:
"This change has allowed tasks to stay at higher priority for much longer so the sleep avg decay of high credit tasks has been changed to match the rate of rise during periods of sleep (which I wanted to do originally but was limited by the first problem). This makes for much more sustained interactivity at extreme loads, and much less X jerkiness."
Read on for Con's full explanation and changelog.
Andrew Morton [interview] has released 2.6.0-test3-mm1, with the following comments:
" - Manfred might have fixed the long-standing task_struct use-after-free bug which happens occasionally on preemptible kernels.
- This kernel immediately triplefaults when compiled with gcc-2.95.3 and CONFIG_KGDB. It is due to compiling with "-ggdb" or "-gdwarf-2". When compiled with "-g" it works OK, but gdb screws that up.
- Moral: use a later gcc if you're a kgdb user.
- Lots of fixes for random things, as usual. Keep 'em coming."
Read on for his full changelog.
Continuing to build toward an official 2.6 kernel, Linux creator Linus Torvalds today released the 2.6.0-test3 kernel, noting, "The bulk of the diff by far is various architecture updates, and in particular bringing MIPS a bit closer to being up-to-date for 2.6.x But there's arm, alpha, h8300 and ia64 updates too."
This kernel release also lays the foundation for a more secure kernel, bringing in Andrew Morton's [interview] recent merge of the SELinux mandatory access control framework. Linus commented briefly, "Merging the SELinux security architecture also ends up growing the patch, even though it may not be all that noticeable for most normal users."
Other additions and fixes include a large collection of CPUFreq updates by Dave Jones [interview], numerous network driver driver updates, a large collection of USB updates, a long-standing PCMCIA lock-up fix, and a read-ahead bug fix. Linus' complete email and changelog follow.
Con Kolivas [interview], who maintains the -ck interactivity patchset for the 2.4 kernel [story] has been maintaining a patchset against the 2.6.0-test kernel tree, also for boosting interactivity. His latest patchset, 'O14int' was released today against both the mainline 2.6.0-test2 development kernel and Andrew Morton's [interview] 2-6.0-test2-mm5 branch. Con describes it simply as "more duct tape interactivity tweaks", going on to detail the recent changes.