On Sat, 2012-06-02 at 23:47 +0100, Ben Hutchings wrote:
> We have about 2 weeks to the freeze in which to enable/backport any
> extra features wanted for wheezy. I have a 4-day weekend now but I am
> not going to spend it all on Debian hacking!
>
> Here's a list of a few to be considered:
>
> - [linux-tools] build more tools:
> - hv (something to do with Hyper-V; requested by MS)
> - usbip (#568362); the current usbip userland package is apparently
> *incompatible* with the usbip driver we build!

The above are now done.

> - lguest (#457652)

I now agree with Bastian and Moritz that we shouldn't enable lguest at
all. i386-on-i386 with no stable PV or management ABI is not very
interesting. kernel-package can build an lguest package, so let people
build both kernel and userland parts of it that way if they want.

> - btrfs improvements (rebalance, general performance, error handling)
> - do these have any dependencies on VFS changes?

Bastian? This now has a bug number, #678797.

> - network teaming driver

This turned out to depend on too many other networking changes, so I
gave up on it.

> - network byte queue limits (anti-bufferbloat)

Still to look at.

> - [armhf] LPAE support

Arnaud?

> - procfs hidepid and hidegid mount options (#669028)

Done.

> - CPU autoloading support
> - discussed in http://bugs.debian.org/654957#20
> - depends on conversion of CPU sysdevs to ordinary devices

Done.

> - fixes for THP with slow writeback (#675493)
> - seccomp filters and no_new_privs (#675615)
> - yama LSM
> - [arm] BPF JIT compiler

Still to look at.

[...]
> - move more maintainer script logic to linux-base

Still to-do.

[...]
> - [armhf] omapdrm driver
> - [x86] gma500 support for new chips
> - [x86] i915 improvements in Ivy Bridge support
> - radeon support for new chips
> - NVMe driver
> - tg3 support for 57766 chip
> - ipheth support for iPhone 4S
> - DRM/KMS driver for DisplayLink
> - qmi_wwan driver (#670241)

Still to-do.

Ben.

--
Ben Hutchings
I say we take off; nuke the site from orbit. It's the only way to be sure.

Ben Hutchings <ben@decadent.org.uk> writes:

> On Sat, 2012-06-02 at 23:47 +0100, Ben Hutchings wrote:
>> We have about 2 weeks to the freeze in which to enable/backport any
>> extra features wanted for wheezy. I have a 4-day weekend now but I am
>> not going to spend it all on Debian hacking!
>>
>> Here's a list of a few to be considered:
>>
>> - [linux-tools] build more tools:
>> - hv (something to do with Hyper-V; requested by MS)
>> - usbip (#568362); the current usbip userland package is apparently
>> *incompatible* with the usbip driver we build!
>
> The above are now done.
>
>> - lguest (#457652)
>
> I now agree with Bastian and Moritz that we shouldn't enable lguest at
> all. i386-on-i386 with no stable PV or management ABI is not very
> interesting. kernel-package can build an lguest package, so let people
> build both kernel and userland parts of it that way if they want.
>
>> - btrfs improvements (rebalance, general performance, error handling)
>> - do these have any dependencies on VFS changes?
>
> Bastian? This now has a bug number, #678797.
>
>> - network teaming driver
>
> This turned out to depend on too many other networking changes, so I
> gave up on it.
>
>> - network byte queue limits (anti-bufferbloat)
>
> Still to look at.
>
>> - [armhf] LPAE support
>
> Arnaud?
>

if one boots a kernel with LPAE enabled on a system not supporting LPAE,
it'll stop booting very early with the "well known message" :

Error: unrecognized/unsupported processor variant

This means it should be enabled only on SoC with LPAE support and afaik,
none of our armhf platforms have it.


>> - procfs hidepid and hidegid mount options (#669028)
>
> Done.
>
>> - CPU autoloading support
>> - discussed in http://bugs.debian.org/654957#20
>> - depends on conversion of CPU sysdevs to ordinary devices
>
> Done.
>
>> - fixes for THP with slow writeback (#675493)
>> - seccomp filters and no_new_privs (#675615)
>> - yama LSM
>> - [arm] BPF JIT compiler
>
> Still to look at.

hm. I thought it was already in 3.2 but looks like it was not. BPF JIT
on arm is limited to commit
ddecdfcea0ae891f782ae853771c867ab51024c2. Will look at that. There are
some kconfig fixes too but I'm not sure we need them.

>
> [...]
>> - move more maintainer script logic to linux-base
>
> Still to-do.
>
> [...]
>> - [armhf] omapdrm driver

still not in mainline.

Arnaud


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 87lijcsm8z.fsf@lebrac.rtp-net.org">http://lists.debian.org/87lijcsm8z.fsf@lebrac.rtp-net.org

Arnaud Patard (Rtp) <arnaud.patard@rtp-net.org> writes:


>> [...]
>>> - [armhf] omapdrm driver
>
> still not in mainline.

replying to self: s/in mainline/out of staging/

Arnaud


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 87hau0s6r5.fsf@lebrac.rtp-net.org">http://lists.debian.org/87hau0s6r5.fsf@lebrac.rtp-net.org

On 2012-06-04 20:42 +0200, Julien Cristau wrote:

> We'll probably also need some nouveau updates to (at least) update to
> the 1.0.0 ABI, I haven't had time to look at this at all, but I think
> Maarten had a backport at
> http://people.canonical.com/~mlankhorst/drm-abi-patches.tgz but doesn't
> have the hw to test them.

We may have a volunteer now (Michele, CC'ed), but the patchset does not
cleanly apply to the 3.2.21 kernel:

,----
| $ tar xf ../drm-abi-patches.tgz -O | patch -Np1 --dry-run
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| patching file drivers/gpu/drm/nouveau/nvd0_display.c
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| Hunk #1 succeeded at 292 (offset 2 lines).
| patching file drivers/gpu/drm/nouveau/nouveau_channel.c
| Hunk #1 succeeded at 425 (offset -9 lines).
| patching file drivers/gpu/drm/nouveau/nouveau_dma.h
| patching file drivers/gpu/drm/nouveau/nouveau_channel.c
| patching file drivers/gpu/drm/nouveau/nouveau_dma.c
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| Hunk #1 succeeded at 1032 (offset 2 lines).
| patching file drivers/gpu/drm/nouveau/nouveau_state.c
| patching file drivers/gpu/drm/nouveau/nouveau_display.c
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| Hunk #1 FAILED at 1672.
| 1 out of 1 hunk FAILED -- saving rejects to file drivers/gpu/drm/nouveau/nouveau_drv.h.rej
| patching file drivers/gpu/drm/nouveau/nvc0_fifo.c
| patching file drivers/gpu/drm/nouveau/nvc0_graph.c
| patching file drivers/gpu/drm/nouveau/nouveau_fence.c
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| Hunk #1 succeeded at 1662 (offset 2 lines).
| patching file drivers/gpu/drm/nouveau/nouveau_fence.c
| patching file drivers/gpu/drm/nouveau/nv50_display.c
| patching file drivers/gpu/drm/nouveau/nouveau_bo.c
| patching file drivers/gpu/drm/nouveau/nouveau_drv.h
| patching file drivers/gpu/drm/nouveau/nouveau_gem.c
`----

Maarten, could you rebase the patches?

Cheers,
Sven


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 87vcidg9mp.fsf@turtle.gmx.de">http://lists.debian.org/87vcidg9mp.fsf@turtle.gmx.de

On 2012-06-26 21:23 +0200, Sven Joachim wrote:

> On 2012-06-04 20:42 +0200, Julien Cristau wrote:
>
>> We'll probably also need some nouveau updates to (at least) update to
>> the 1.0.0 ABI, I haven't had time to look at this at all, but I think
>> Maarten had a backport at
>> http://people.canonical.com/~mlankhorst/drm-abi-patches.tgz but doesn't
>> have the hw to test them.
>
> We may have a volunteer now (Michele, CC'ed), but the patchset does not
> cleanly apply to the 3.2.21 kernel:
>
> ,----
> | $ tar xf ../drm-abi-patches.tgz -O | patch -Np1 --dry-run
> | [...]
> `----

Scrap that please, this is not the right way to apply the patches. The
following commands should actually work (assuming you're on sid and have
the latest 3.2 kernel installed):

sudo apt-get install linux-source-3.2
tar xf /usr/src/linux-source-3.2.tar.bz2
wget http://people.canonical.com/~mlankhorst/drm-abi-patches.tgz
cd linux-source-3.2
tar xf ../drm-abi-patches.tgz
for patch in patches/*; do patch -Np1 -i $patch;done
cp /boot/config-3.2.0-3-amd64 .config
make deb-pkg

Then there should be a linux-image-3.2.21-<something>.deb in the parent
directory that you can install with dpkg. Michele, could you do that?

Cheers,
Sven


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 87lij7ijru.fsf@turtle.gmx.de">http://lists.debian.org/87lij7ijru.fsf@turtle.gmx.de

On 2012-06-28 16:38 +0200, Sven Joachim wrote:

> Scrap that please, this is not the right way to apply the patches. The
> following commands should actually work (assuming you're on sid and have
> the latest 3.2 kernel installed):
>
> sudo apt-get install linux-source-3.2
> tar xf /usr/src/linux-source-3.2.tar.bz2
> wget http://people.canonical.com/~mlankhorst/drm-abi-patches.tgz
> cd linux-source-3.2
> tar xf ../drm-abi-patches.tgz
> for patch in patches/*; do patch -Np1 -i $patch;done
> cp /boot/config-3.2.0-3-amd64 .config
> make deb-pkg

Michele replied in private mail that the patched kernel works fine with
3D acceleration.

Cheers,
Sven


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 878vf7i822.fsf@turtle.gmx.de">http://lists.debian.org/878vf7i822.fsf@turtle.gmx.de

On Sun, 2012-06-24 at 15:43 +0100, Ben Hutchings wrote:
> On Sat, 2012-06-02 at 23:47 +0100, Ben Hutchings wrote:
[...]
> > - btrfs improvements (rebalance, general performance, error handling)
> > - do these have any dependencies on VFS changes?
>
> Bastian? This now has a bug number, #678797.

Still waiting on this.

[...]
> > - network byte queue limits (anti-bufferbloat)
>
> Still to look at.

I prepared a backport, but it includes an unavoidable ABI change. We
need to find a good time to do this.

[...]
> > - fixes for THP with slow writeback (#675493)

These should be in Linux 3.2.25. However, they appear to include an
unavoidable ABI change.

> > - seccomp filters and no_new_privs (#675615)

Are these still worthwhile for wheezy? We're probably not going to have
any packaged applications that can use it, although libseccomp is
included.

> > - yama LSM

Still to look at.

> > - [arm] BPF JIT compiler

Arnaud has done this.

> [...]
> > - move more maintainer script logic to linux-base
>
> Still to-do.
>
> [...]
> > - [armhf] omapdrm driver
> > - [x86] gma500 support for new chips
> > - [x86] i915 improvements in Ivy Bridge support
> > - radeon support for new chips
> > - NVMe driver
> > - tg3 support for 57766 chip

Still to do.

> > - ipheth support for iPhone 4S

I missed it at the item, but that went into 3.2.9.

> > - DRM/KMS driver for DisplayLink

Still to do.

> > - qmi_wwan driver (#670241)

This turned out not to be needed.

Time for some fresh lists rather than continuing to indent the old one
further.

Features, may not be allowed now:

- network byte queue limits (anti-bufferbloat) [ABI change]
- seccomp filters and no_new_privs (#675615)
- yama LSM

Bug fixes and hardware support, probably still allowed:

- btrfs improvements (rebalance, general performance, error handling)
- fixes for THP with slow writeback (#675493) [ABI change]
- [armhf] omapdrm driver
- [x86] gma500 support for new chips
- [x86] i915 improvements in Ivy Bridge support
- radeon support for new chips
- NVMe driver
- tg3 support for 57766 chip
- DRM/KMS driver for DisplayLink

Ben.

--
Ben Hutchings
Man invented language to satisfy his deep need to complain. - Lily Tomlin

Hi,

Ben Hutchings wrote:
>> On Sat, 2012-06-02 at 23:47 +0100, Ben Hutchings wrote:

>>> - seccomp filters and no_new_privs (#675615)
>
> Are these still worthwhile for wheezy? We're probably not going to have
> any packaged applications that can use it, although libseccomp is
> included.

chromium-browser in wheezy uses the kernel interface[*]. I think it's
worthwhile.

Thanks,
Jonathan
[*] Noticed because people using 3.5-rc kernels were running into the
bug fixed by 5651721edec2: http://bugs.debian.org/679827


--
To UNSUBSCRIBE, email to debian-kernel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: http://lists.debian.org/20120729225907.GA21617@burratino