LWN: Comments on "Some numbers and thoughts on the stable kernels"

Some numbers and thoughts on the stable kernels

gvy — Thu, 16 Aug 2012 13:20:17 +0000

For the record, that bugfix has been discussed and accepted within two weeks of moving the bugreport to LKML and within something like 8 hours of wall clock. A lil' more patience ;-)

Some numbers and thoughts on the stable kernels

gvy — Sat, 11 Sep 2010 11:01:17 +0000

> "volunteers are still our leading source of fixes"
Well, I tried to get the patch for deadlock issue our folks have nailed down and fixed (it was rather worked around upstream) along at least into 2.6.27.y since this summer, and so far rather failed to bring attention to the issue:

https://bugzilla.kernel.org/show_bug.cgi?id=15658

Tried to forward the message with brief explanation to LKML on Aug 28 but seems like it didn't pass through (at least not to lkml.org archive).

I've re-read all of tux.org/lkml just in case, but still need advice on *how* volunteers should pass the fixes for such non-obvious issues?

Thanks anyone who might find some time to aid with this one.

Some numbers and thoughts on the stable kernels

lacostej — Mon, 06 Sep 2010 09:29:36 +0000

Any reason why Adrian has 'left' the community?

I loved to see his dedication. Has someone taken his role?

Some numbers and thoughts on the stable kernels

hingo — Tue, 31 Aug 2010 21:16:17 +0000

I've been playing with the idea of producing similar stats for MySQL/MariaDB development, to benchmark it against Linux and other similar projects. This is an excellent idea to add to the analysis! Thanks.

Some numbers and thoughts on the stable kernels

bfields — Tue, 31 Aug 2010 16:00:08 +0000

If I read lwn at work, am I shirking? If I read it after hours, am I bringing my work home with me? Help!

Some numbers and thoughts on the stable kernels

macson_g — Tue, 31 Aug 2010 12:17:36 +0000

Yep: http://lwn.net/Articles/402327/

Some numbers and thoughts on the stable kernels

njd27 — Tue, 31 Aug 2010 12:00:27 +0000

Are there any signs of him going crazy yet?

Some numbers and thoughts on the stable kernels

bfields — Mon, 30 Aug 2010 22:32:02 +0000

Hm, a simple "Signed-off-by:" should suffice.

Yeah. Here's a previous discussion, where Linus says:

So when you save a patch from oblivion by passing it on to the right person, and get it submitted when it was originally dropped by some reason, you're actually doing a fundamentally important job. Maybe it's just one small piece of the puzzle, but hey, you'd only get one small line in the changeset, so the credit (or blame ;) really is appropriate.

Some numbers and thoughts on the stable kernels

ironiridis — Mon, 30 Aug 2010 16:57:36 +0000

With regard to "never" seeing someone call GKH a dickhead, you might have comments from Hknr filtered on LWN.

Filtering ENABLED

mebrown — Mon, 30 Aug 2010 14:49:12 +0000

Nice! Thanks!

Some numbers and thoughts on the stable kernels

spaetz — Mon, 30 Aug 2010 08:04:22 +0000

>Who is ever really off the clock in this field?

As a side note, I have seen some research on commit patterns across projects and it was really entertaining to see some projects following strict Mo-Fri 9-5 patterns while others (also including commercial developers) committed all over the place including weekends, nights etc.

I would love to see that research across a wide range

Some numbers and thoughts on the stable kernels

eduard.munteanu — Sun, 29 Aug 2010 14:44:31 +0000

Hm, a simple "Signed-off-by:" should suffice. It's already standard practice for people who maintain their own tree and relay stuff upstream, and usually those trees are rebased often. So it might make sense to use that for singular cases as well (even if they're not merges).

This...

Signed-off-by: <original author>
Signed-off-by: <maintainer>
Signed-off-by: Linus Torvalds

Could turn into this...

Signed-off-by: <original author>
Signed-off-by: <spotter>
Signed-off-by: Greg KH

Possibly with some variation, such as whether or not you keep the maintainer in the Signed-off-by chain.

Filtering

zooko — Sun, 29 Aug 2010 14:03:10 +0000

Hey thanks, it works nicely!

Some numbers and thoughts on the stable kernels

maks — Sun, 29 Aug 2010 13:05:36 +0000

Thanks josh for clearing up and your nice comment. :)

As one can see on aboves authorship table Ben Hutchings is doing an incredible great job in the last year.

Some numbers and thoughts on the stable kernels

maks — Sun, 29 Aug 2010 13:00:17 +0000

linux-base is used for the UUID transition. In Ubuntu it was the udev package that renamed the fstab and some bootloaders. afais linux-base is in experimental. When installing from exp one needs to tell apt to do so with ``-t'' switch aka:
apt-get install linux-image-2.6.35-trunk-amd64 -t experimental

Hope that helps otherwise please ask on debian-kernel mailinglist.

Some numbers and thoughts on the stable kernels

maks — Sun, 29 Aug 2010 12:56:03 +0000

Ok thanks for the explanations.

Got confused by the title of the table "Most active stable contributors" and thought it was about the ones that notify stable. Could then be renamed to "Most active stable authors".

Indeed there is currently no possibility to easily see who forwarded the patches to gregkh.

Some numbers and thoughts on the stable kernels

erwbgy — Sun, 29 Aug 2010 09:36:53 +0000

Well, that may be true, but he is definitely a more useful dickhead than you are.

Filtering

corbet — Sun, 29 Aug 2010 03:44:16 +0000

It's in the My Account area.

Some numbers and thoughts on the stable kernels

zooko — Sun, 29 Aug 2010 02:56:44 +0000

Hey where is that tool for filtering the comments of certain posters out of your view of LWN?

Some numbers and thoughts on the stable kernels

hmh — Sun, 29 Aug 2010 00:46:12 +0000

Yes, that would be nice.

Also, when there is no "spotted-by:" but there is a "cc: stable", it likely means that the patch author noticed it should go into stable.

That said, I think "stable-proposal-by" would be a better tag.

Some numbers and thoughts on the stable kernels

ncm — Sat, 28 Aug 2010 23:44:39 +0000

Yes, thanks Maks! I just loaded linux-image-2.6.35-1~experimental.2 on a Dell Latitude D430 sid and it fixed the suspend/resume problems. (Q though: why does it depend on a linux-base that doesn't exist?)

Some numbers and thoughts on the stable kernels

ncm — Sat, 28 Aug 2010 20:56:24 +0000

Now I know where to send Hknr messages.

Some numbers and thoughts on the stable kernels

corbet — Sat, 28 Aug 2010 19:57:33 +0000

I agree it would be nice to credit people who do that kind of work. There is really no information trail at the moment which would make that possible, though. Maybe we need a Spotted-by: tag to mark patches directed to stable by people other than their author?

Some numbers and thoughts on the stable kernels

josh — Sat, 28 Aug 2010 19:46:54 +0000

I think maks (AKA maximilian attems, one of the Debian kernel maintainers) is suggesting that in addition to crediting the author of the patch, for stable patches it seems appropriate to note the person who dug up a patch from mainline that fixed a particular bug and passed it along to stable@ for inclusion in a stable kernel, even if they didn't *write* the mainline patch. And I'd agree with that; while not the same metric as counting authors, it seems a useful metric for stable kernels.

On that note, thanks to maks and the other Debian kernel maintainers for doing an awesome job. :)

Some numbers and thoughts on the stable kernels

corbet — Sat, 28 Aug 2010 19:16:06 +0000

I simply look at the author of each patch as is shown in the stable git repository.

A quick look through 2.6.32.y shows you CC'd on a number of patches, but you are not the author of them in stable or mainline. Am I missing something?

Some numbers and thoughts on the stable kernels

maks — Sat, 28 Aug 2010 19:01:05 +0000

Your stats concerning the personal contribution to stable seem at first look pretty screwed:

First of all it is not easy to find the one that forwarded a particular patch to stable.
Secondly most in all times this random person gets gets showed into the Cc: of the stable patch and thus is not distinguisable from usual Cc's that happen to be listed there.

A quick grep in the stable queue 2.6.32 releases showed 47 contributions of mine, but I'm 100% sure not to be the top contributor of it. But that forwarded patch number is certainly a lower bound.

Please explain how you generated the stats concerning the most active contributor. Thanks.

Some numbers and thoughts on the stable kernels

BenHutchings — Sat, 28 Aug 2010 15:00:33 +0000

We discussed this at the LPC last year. If I remember correctly, Greg's answer was that:
1. Distributions need to let him know which stable releases they want to keep going.
2. They need to help in gathering updates, and potentially to take over maintenance.

unpolite, unrespectful, uninformative and subjective

dmk — Sat, 28 Aug 2010 13:07:15 +0000

From the LWN Comment editor:
<quote>Please try to be polite, respectful, and informative, and to provide a useful subject line.</quote>

Some numbers and thoughts on the stable kernels

dambacher — Sat, 28 Aug 2010 07:44:36 +0000

What I miss in the figures is the duration (e.g. in weeks) a kernel was maintained -> wich was a long time kernel and wich one was really rotten

Some numbers and thoughts on the stable kernels

Hknr — Sat, 28 Aug 2010 07:17:33 +0000

Well, Greg KH *is* a dickhead, no question about it.
He's one of the few kernel guys whose messages go straight
to /dev/null in my LKML filter. This also lowers the
traffic by a considerable amount as you no longer see
his endless "stable preview" threads.

Some numbers and thoughts on the stable kernels

dlang — Sat, 28 Aug 2010 01:17:14 +0000

this is where it gets interesting.

it can be that someone just volunteers to maintain it after Greg finishes with it, it can be that one (or more in the case of ) major distros base their release kernel off of it, and as a result someone takes on the maintenance of the kernel after it would finish the normal process.

about the only criteria is an elimination criteria. Namely that the kernel in question must not have a disastrous amount of regressions discovered in the normal -stable period.

Some numbers and thoughts on the stable kernels

mpr22 — Sat, 28 Aug 2010 00:04:33 +0000

People almost never call anyone a dickhead on LWN :)

Some numbers and thoughts on the stable kernels

nicooo — Fri, 27 Aug 2010 23:42:44 +0000

What are the criteria for deciding which kernels will be supported? (e.g. what makes 2.6.27 different from .26 or .28)

Some numbers and thoughts on the stable kernels

corbet — Fri, 27 Aug 2010 21:59:58 +0000

Who is ever really off the clock in this field?

A patch posted from a company email address is assumed to be posted with that company's blessing. Anybody who doesn't have that blessing is probably violating all kinds of rules. In short: when somebody posts a patch, in the absence of information to the contrary, we attribute it to their employer. The results are necessarily approximate, but I'm not sure how to do them better.

Some numbers and thoughts on the stable kernels

corbet — Fri, 27 Aug 2010 21:57:06 +0000

Is there some context in which Adrian should have been mentioned? I didn't talk about 2.6.16, really...and he hasn't really been present in the kernel community for a while now.

Some numbers and thoughts on the stable kernels

drag — Fri, 27 Aug 2010 21:55:42 +0000

The amount of contributions, by lines of code or by patches, does not translate directly to bug fixes. Most of them are probably features and drivers.

Some numbers and thoughts on the stable kernels

spender — Fri, 27 Aug 2010 21:44:17 +0000

Third question: is there a reason why there is no mention of Adrian Bunk in the article?

-Brad

Some numbers and thoughts on the stable kernels

jspaleta — Fri, 27 Aug 2010 21:42:35 +0000

I wouldn't make the assumption about how corporate involvement is ascertained. I believe the author can go into detail as to how that is done if you need it clarified.

-jef

Some numbers and thoughts on the stable kernels

spender — Fri, 27 Aug 2010 21:21:25 +0000

"volunteers are still our leading source of fixes"

Though volunteers beat out any other single company source of fixes, if I'm interpreting the chart correctly for 2.6.32, (100% - 15.3%) = 84.7% are coming from companies.

That said, I can imagine a good amount of the patches being submitted from company email addresses (which is what I assume the stats are generated off of) are done when the employee is "off the clock."

So my question: do we (and how do we) actually know how much of the kernel development is really a volunteer effort?

BTW, I've never seen anyone on this site call Greg a "dickhead" or anything close to it (I did a search to check, feel free to do your own). Some people, myself included, disagree with how security is conveyed and sometimes handled, but it's a gross mischaracterization to equate that with childish ad-hominem attacks that were never uttered.

I appreciate the effort Greg puts into the stable releases -- it's certainly a lot of work. As for the major "enterprise" distributions that benefit from Greg's work, do we know what else they could be doing in addition to contributing fixes to lessen the burden?

-Brad

Some numbers and thoughts on the stable kernels

flewellyn — Fri, 27 Aug 2010 20:38:45 +0000

Linus was certainly right about one thing: there are indeed people who claim that Greg KH is a dickhead and worse, on a weekly basis. Such comments are as productive as one might expect.