kernel 2.7 when?

danbond_98 · n00b Joined: 23 Aug 2003 Posts: 8 Location: Exeter, UK

Any suggestions, when's it gonna start being developed or am i missing something huge and has work started on it already?

hifi · Apprentice Joined: 17 Jul 2002 Posts: 184 Location: Graz

Helena · Posted: Mon Jan 05, 2004 1:20 pm Post subject: Re: kernel 2.7 when?

ewan.paton · Posted: Mon Jan 05, 2004 1:43 pm Post subject:

they will probably wait a bit to keep the talented coders working on bugfixes etc, in fairness once 2.4 got the preemtable patch i was fine, what are you wanting for 2.7 as there are patches for most stuff unless you are waiting to scale to insaine amounts of cpus
_________________
Giay tay nam | Giay nam cao cap | Giay luoi

Stormy Eyes · Posted: Mon Jan 05, 2004 1:52 pm Post subject: Re: kernel 2.7 when?

shm · Posted: Mon Jan 05, 2004 2:34 pm Post subject:

I think there was about 7 or 8 stable releases of 2.4 before 2.5.0 was released, perhaps we'll see the same thing.
_________________
what up

NME · Posted: Mon Jan 05, 2004 3:22 pm Post subject:

another thing: what can we expect from the new dev series? what will be the focus of the 2.7 tree?
_________________
f# that.

Stormy Eyes · Posted: Mon Jan 05, 2004 3:29 pm Post subject:

steel300 · Veteran Joined: 10 Jul 2003 Posts: 1155

Stormy Eyes · Posted: Mon Jan 05, 2004 4:29 pm Post subject:

steel300 · Veteran Joined: 10 Jul 2003 Posts: 1155

Yes 2.5 was a complete rewrite of 2.4. The basic concepts of semaphores and spin locks were still there, but other than that, it's a kernel from scratch. I can't say how often Linus does complete rewrites. The kernel can run on any number of processors up to 255. Performance degrades greatly that high though. Problems with memory access and interrupts cause many issues with more than one processor. Since NUMA is a relatively newly implemented concept, it's support isn't that great. NUMA (Non-Uniform Memory Access) brings along its own issues.

To push that Opteron rig, run seti@home to eat extra cycles. I've got a dual p3-933 with both procs always walled.
_________________
Rationality is the recognition of the fact that nothing can alter the truth and nothing can take precedence over that act of perceiving it.

Stormy Eyes · Posted: Mon Jan 05, 2004 4:48 pm Post subject:

fathergrief · n00b Joined: 04 Oct 2002 Posts: 35 Location: Alaska

CharlieS · Posted: Tue Jan 06, 2004 3:31 am Post subject:

2.6 has only been out for like a month.. hold your horses...

Stormy Eyes · Posted: Tue Jan 06, 2004 3:53 am Post subject:

Nylle · Posted: Tue Jan 06, 2004 1:30 pm Post subject:

ewan.paton · Posted: Tue Jan 06, 2004 3:39 pm Post subject:

is scsi performance really an issuse, i've never had much probs whith it and since high end stuff is going fibre channel[1] and lowend will probably go serial ata, i hope the put most work into scalling the cpus with the sony/ibm cell chips out in 2005ish

[1] mm 2 gigs drive bandwidth
_________________
Giay tay nam | Giay nam cao cap | Giay luoi

Stormy Eyes · Posted: Tue Jan 06, 2004 4:03 pm Post subject:

Kihaji · Apprentice Joined: 12 Sep 2002 Posts: 230

Stormy Eyes · Posted: Tue Jan 06, 2004 4:32 pm Post subject:

steel300 · Veteran Joined: 10 Jul 2003 Posts: 1155

SCSI support won't go away anytime soon. It is still heavily used in large servers. Dropping SCSI support would only hurt linux integration in areas where we already own a majority of the market share.
_________________
Rationality is the recognition of the fact that nothing can alter the truth and nothing can take precedence over that act of perceiving it.

ewan.paton · Posted: Tue Jan 06, 2004 5:26 pm Post subject:

i wasnt saying we should drop scsi suport, mostly just wondering what was wrong with it at the moment
_________________
Giay tay nam | Giay nam cao cap | Giay luoi

steel300 · Veteran Joined: 10 Jul 2003 Posts: 1155

SCSI, by original design, is slow to stabilize. It takes longer to stabilize a scsi drive for use, than it does an IDE drive. It has always been that way. That's the main problem with SCSI. Other than that, it pretty much works.
_________________
Rationality is the recognition of the fact that nothing can alter the truth and nothing can take precedence over that act of perceiving it.

Nylle · Posted: Wed Jan 07, 2004 12:16 am Post subject:

If I remember correctly, it's just that the SCSI subsystem is quite messy, and messy is not good when it comes to the kernel. A cleaner implementation leads to code that is easier to debug and nicer driver interfaces.

The IDE-subsystem has been a mess too, and there was a lot of debate about how to re-implement it. It's very sticky stuff, that affects many other parts of the kernel, and quite complicated to write. They dumped an initial complete re-design, and went with a less invasive one if I'm not mistaken.

If you want to follow kernel development, read kernel traffic, it is a great compilation on what has gone on in the dev mailing lists over the week.
_________________
"Do you hear that sound your Highness?"
"Those are the shrieking eels, they always grow louder when they are about to feed on human flesh."