Skip to contentSkip to conversation

Ubuntu patch sync

PATCH In review Ben Collins 9 participants 13 messages Jan 4, 2006 – Jan 5, 2006

lore
Share & tools
lore
#0 lore raw

cover Ubuntu patch sync

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

You can pull from:

master.kernel.org:/pub/scm/linux/kernel/git/bcollins/ubuntu-2.6.git ubuntu-fixes

Or:

rsync://rsync.kernel.org/pub/scm/linux/kernel/git/bcollins/ubuntu-2.6.git ubuntu-fixes

diffstat:

 arch/i386/kernel/acpi/boot.c                 |   10 ++
 arch/i386/kernel/acpi/boot.c                 |   10 ++
 arch/i386/kernel/cpu/cpufreq/p4-clockmod.c   |   10 ++
 arch/i386/kernel/cpu/cpufreq/powernow-k7.c   |   11 +-
 arch/i386/kernel/reboot.c                    |    9 ++
 arch/i386/pci/fixup.c                        |    8 +-
 drivers/acpi/asus_acpi.c                     |   11 +-
 drivers/acpi/processor_idle.c                |   68 +++++++++++++----
 drivers/char/sonypi.c                        |   46 +++++++++++-
 drivers/input/misc/wistron_btns.c            |   20 ++++-
 drivers/input/serio/i8042-x86ia64io.h        |    9 ++
 drivers/macintosh/therm_adt746x.c            |   40 ++++++----
 drivers/macintosh/via-pmu.c                  |    5 +
 drivers/net/irda/nsc-ircc.c                  |  103 +++++++++++++++++++++++++--
 kernel/workqueue.c                           |    6 +
 scripts/kconfig/conf.c                       |   19 ++++
 16 files changed, 323 insertions(+), 62 deletions(-)
#1 lore raw
Quoted 4 lines · On Wed, 04 Jan 2006 16:59:02 -0500 Ben Collins <[email protected]> wrote:

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Why not submit them too?

Signature · 2 lines
Stephen Hemminger <[email protected]>
OSDL http://developer.osdl.org/~shemminger
#2 lore raw
Quoted 9 lines · Stephen Hemminger
Quoted 4 lines · On Wed, 04 Jan 2006 16:59:02 -0500 Ben Collins <[email protected]> wrote:

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Why not submit them too?

Because neither I nor Ubuntu maintains them as upstream. I would rather leave it to the upstream authors of those drivers (e.g. rt2400, rt2500, unionfs, etc.) to submit their own code to Linus.

Signature · 3 lines
   Ben Collins <[email protected]>
   Developer
   Ubuntu Linux
#3 lore raw
Quoted 9 lines · On Wed, 04 Jan 2006 17:12:48 -0500 Ben Collins <[email protected]> wrote:
Quoted 8 lines · Stephen Hemminger
Quoted 4 lines · On Wed, 04 Jan 2006 16:59:02 -0500 Ben Collins <[email protected]> wrote:

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Quoted 6 lines
Quoted 1 line

Why not submit them too?

Because neither I nor Ubuntu maintains them as upstream. I would rather leave it to the upstream authors of those drivers (e.g. rt2400, rt2500, unionfs, etc.) to submit their own code to Linus.

Just want to needle the upstream drivers to submit and get reviewed.

Signature · 2 lines
Stephen Hemminger <[email protected]>
OSDL http://developer.osdl.org/~shemminger
#4 lore raw
Quoted 12 lines · Stephen Hemminger
Quoted 9 lines · On Wed, 04 Jan 2006 17:12:48 -0500 Ben Collins <[email protected]> wrote:
Quoted 8 lines · Stephen Hemminger
Quoted 4 lines · On Wed, 04 Jan 2006 16:59:02 -0500 Ben Collins <[email protected]> wrote:

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

Why not submit them too?

Because neither I nor Ubuntu maintains them as upstream. I would rather leave it to the upstream authors of those drivers (e.g. rt2400, rt2500, unionfs, etc.) to submit their own code to Linus.

Just want to needle the upstream drivers to submit and get reviewed.

After dealing with literally dozens of upstream drivers, I think the reasons boil down to a few categories:

1 - They know their code is crappy and they don't want to deal with code review.

2 - They want to retain total control of their code. Having it in the kernel tree means that the driver can be modified by others (in usually correct ways) without their consent. They don't want to have to track these changes and accommodate them. Also, external drivers tend to retain a lot of backward compatibility for older kernel versions, that would surely be dropped if it was placed in the kernel proper.

3 - They don't think their driver is important enough. They have a small user base, and deal directly with those users.

4 - They just don't know how.

Not defending any of these reasons. I'd love to not have all this work of pulling in and tracking the drivers that our users need/want, but it's going to be a lot of work. Maybe I'll start emailing them about getting their code in the kernel tree.

Signature · 3 lines
   Ben Collins <[email protected]>
   Developer
   Ubuntu Linux
#5 lore raw
Quoted 20 lines · On Wed, 04 Jan 2006 17:45:39 -0500 Ben Collins <[email protected]> wrote:
Quoted 19 lines · Stephen Hemminger
Quoted 15 lines · On Wed, 04 Jan 2006 17:12:48 -0500 Ben Collins <[email protected]> wrote:
Quoted 9 lines · Stephen Hemminger
Quoted 4 lines · On Wed, 04 Jan 2006 16:59:02 -0500 Ben Collins <[email protected]> wrote:

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Why not submit them too?

Because neither I nor Ubuntu maintains them as upstream. I would rather leave it to the upstream authors of those drivers (e.g. rt2400, rt2500, unionfs, etc.) to submit their own code to Linus.

Quoted 7 lines
Quoted 1 line

Just want to needle the upstream drivers to submit and get reviewed.

After dealing with literally dozens of upstream drivers, I think the reasons boil down to a few categories:

1 - They know their code is crappy and they don't want to deal with code review.

Then find some friendly reviewers. It doesn't have to be a "throw crap over the wall and see what flys back review". If it has dicey locking and race issues, then the security and reliability of the whole system is effected.

Quoted 6 lines

2 - They want to retain total control of their code. Having it in the kernel tree means that the driver can be modified by others (in usually correct ways) without their consent. They don't want to have to track these changes and accommodate them. Also, external drivers tend to retain a lot of backward compatibility for older kernel versions, that would surely be dropped if it was placed in the kernel proper.

And therefore their driver gets broken regularly. The modified by others is almost always for small compatibility stuff.

Quoted 2 lines

3 - They don't think their driver is important enough. They have a small user base, and deal directly with those users.

If it is important enough for Ubuntu to ship, then it probably is important enough to get out of the closet.

Quoted 1 line

4 - They just don't know how.

Which is true, unfortunately, unless they read about it on a regular basis, most people will plead ignorance.

Quoted 4 lines

Not defending any of these reasons. I'd love to not have all this work of pulling in and tracking the drivers that our users need/want, but it's going to be a lot of work. Maybe I'll start emailing them about getting their code in the kernel tree.

I don't mean to play shoot the messenger, you are helping a lot by doing this I wish every distro did.

Signature · 2 lines
Stephen Hemminger <[email protected]>
OSDL http://developer.osdl.org/~shemminger
#6 lore raw
Quoted 7 lines · Stephen Hemminger
Quoted 4 lines

Not defending any of these reasons. I'd love to not have all this work of pulling in and tracking the drivers that our users need/want, but it's going to be a lot of work. Maybe I'll start emailing them about getting their code in the kernel tree.

I don't mean to play shoot the messenger, you are helping a lot by doing this I wish every distro did.

FWIW, Red Hat has been under constant pressure from users asking us to merge driver x 'because $otherdistro has it' for a long time.

The first reason we say no 99.9% of the time is because to stay in a position where we can quickly rebase to a new upstream release, you have to be light enough to not have to spend a day rediffing/fixing add-on patches.

The second reason (and this is actually more important than the first, but as I'm a selfish sod, I thought of my own wellbeing first), is that historically there have been some vendors who have taken the position "Our driver is in Red Hat's distro, we don't care about anyone else". For obvious reasons, this isn't a good thing, and something we now go out of our way to discourage.

Quoted 1 line

From time to time, we do carry drivers that aren't upstream yet.

ipw2x00 got a good beating in RHEL4 & FC3 before it made it into Linus' tree. A lot of issues were found in that driver early on by rhel beta-testers, which Intel went and fixed, long before it was proposed for upstream inclusion. Whilst some of those issues would probably have been found by -mm exposure eventually, living in a distro tree gets it a lot more exposure.

Getting 'previews' of drivers into the grubby hands of end-users who are prepared for a bumpy ride, and are prepared to give feedback to improve things is invaluable.

For Fedora at least, I don't object to the notion of carrying a handful of continually improving drivers which are making progress at upstream inclusion. What I don't want to entertain is the idea of grabbing random drivers and dumping them in the distro tree just because their upstream has no motivation at getting them merged in Linus' tree. (Which is sadly a category quite a few of the requests that we get
fall into)

Dave

#7 lore raw
In reply to Ben Collins
Quoted 4 lines · Ben Collins

After dealing with literally dozens of upstream drivers, I think the reasons boil down to a few categories:

<...>

You missed one:

5 - They've implemented Yet Another IEEE802.11 stack instead of embracing and extending the Intel one we already have, and are hence urinating into the atmospheric disturbance.

That's one of the reasons why I merged bcm43xx and softmac into the Fedora kernel and none of the others, FWIW -- Johannes is actually working on improving what we have in the kernel, rather than just saying "You have to throw it all away because $MYSTACK is better". So I figure softmac has a _much_ better chance of going upstream, even if its feature list isn't quite as comprehensive yet.

http://softmac.sipsolutions.net/softmac-2.6.git

Signature · 1 line
dwmw2
#8 lore raw
Quoted 12 lines · David Woodhouse
Quoted 4 lines · Ben Collins

After dealing with literally dozens of upstream drivers, I think the reasons boil down to a few categories:

<...>

You missed one:

5 - They've implemented Yet Another IEEE802.11 stack instead of embracing and extending the Intel one we already have, and are hence urinating into the atmospheric disturbance.

Quoted 6 lines

That's one of the reasons why I merged bcm43xx and softmac into the Fedora kernel and none of the others, FWIW -- Johannes is actually working on improving what we have in the kernel, rather than just saying "You have to throw it all away because $MYSTACK is better". So I figure softmac has a _much_ better chance of going upstream, even if its feature list isn't quite as comprehensive yet.

Yeah, I did the same for bcm43xx (it's in the ubuntu tree, and working on my powerbook right now with wep :)

Didn't notice softmac had a git tree now. And here I am pulling from mercurial and hand merging :/

Quoted 1 line

http://softmac.sipsolutions.net/softmac-2.6.git

Signature · 3 lines
   Ben Collins <[email protected]>
   Developer
   Ubuntu Linux
#9 lore raw
In reply to Ben Collins
Quoted 10 lines
Quoted 6 lines
Quoted 4 lines

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Why not submit them too?

Because neither I nor Ubuntu maintains them as upstream. I would rather leave it to the upstream authors of those drivers (e.g. rt2400, rt2500, unionfs, etc.) to submit their own code to Linus.

The maximum where unionfs should IMO go ATM is -mm, because it still has refcount issues and occassional crashes.

rt2400/etc is a bit more tricky. There exist several versions and they all have different levels of evolution :-/
Group 1:
- rt* from ralink.com.tw
Group 2:
- rt2400,rt2500,rt2570 "1.1.0" from sf.net/projects/rt2400
- rt2x00 "2.0", also sf.net/projects/rt2400

Nevertheless, I do have rt* 1.1 and 2.0, and unionfs in my patchset...

Jan Engelhardt

#10 lore raw

Jan Engelhardt <[email protected]> :
[...]

Quoted 4 lines

rt2400/etc is a bit more tricky. There exist several versions and they all have different levels of evolution :-/
Group 1:
- rt* from ralink.com.tw

Too much work to merge imho.

Quoted 3 lines

Group 2:
- rt2400,rt2500,rt2570 "1.1.0" from sf.net/projects/rt2400
- rt2x00 "2.0", also sf.net/projects/rt2400

Afaiks it does not apply to the rt2x00 drivers. However these are still strongly different from the ideal in-kernel driver. CodingStyle (arguable) and locking - broken - would need more work. This is not yet another 80211 subsystem but it is not clear if the authors/maintainers plan to merge their stuff in a near future.

--
Ueimor

#11 lore raw
Quoted 9 lines · Francois Romieu
Quoted 3 lines

Group 2:
- rt2400,rt2500,rt2570 "1.1.0" from sf.net/projects/rt2400
- rt2x00 "2.0", also sf.net/projects/rt2400

Afaiks it does not apply to the rt2x00 drivers. However these are still strongly different from the ideal in-kernel driver. CodingStyle (arguable) and locking - broken - would need more work. This is not yet another 80211 subsystem

Currently this is true, but AFAIK they intend to switch to the DeviceScape stack right now (between rt2x00 2.0b3 and 2.0b4).

Dominik

#12 lore raw
In reply to Ben Collins
Quoted 4 lines · Ben Collins

These patches are just attempts to merge code from the ubuntu kernel tree. This is most of the differences between our tree and stock code (not necessarily all differences, since we do have a lot of external drivers pulled in).

Hi Ben,
Just a note in case you don't receive feedbacks to all the patches, it looks like you didn't CC the relevant maintainers.

Anyway, i really appreciate that you try to keep as in sync as possible the Ubuntu kernel to the vanilla.

Ciao,
--
Paolo